Poster Session Schedule

Monday, September 14, 5:20-6:05 EDT

Biomedical applications, from disease to exposome

"A distinctive metabolic profile between Cystic Fibrosis mutational sub-classes and lung function" (3)


Afshan Masood, Proteomics Resource Unit, Obesity Research Center, College of Medicine, King Saud University, PO. Box 2925 (98), Riyadh 11461, Saudi Arabia

Minnie Jacob, King Faisal Specialist Hospital and Research Center

Xinyun Gu, Department of Chemistry, University of Alberta, Edmonton, AB T6G 2R3, Canada

Mai Abdel Jabar, King Faisal Specialist Hospital and Research Center

Hicham Benabdelkamel, Proteomics Resource Unit, Obesity Research Center, College of Medicine, King Saud University, PO. Box 2925 (98), Riyadh 11461, Saudi Arabia

Imran Nizami, Lung Transplant Section, Organ Transplant Center, King Faisal Specialist Hospital and Research Center, Zahrawi Street, Al Maather, Riyadh 11211, Saudi Arabia.

Liang Li, Department of Chemistry, University of Alberta, Edmonton, AB T6G 2R3, Canada

Majed Dasouki, King Faisal Specialist Hospital and Research Center

Anas Abdel Rahman, King Faisal Specialist Hospital and Research Center (Primary Presenter)


Introduction: Cystic fibrosis (CF) is a lethal multisystemic disease having a monogenic origin with numerous mutations. Functional defects in the cystic fibrosis transmembrane conductance receptor (CFTR) protein based on the mutations characteristically divide into distinct classes that have a different metabolic role in the body. Objectives: Previous metabolomics studies compared serum or plasma, exhaled breath, and lung secretion profiles in CF and non-CF patients or with exacerbation of CF. The present study aims to create a comprehensive profile of metabolites altered with CF in general and that of CF Class III and IV, in addition to their lung function. Methods: Herein, we used chemical isotope labeling liquid chromatography-mass spectrometry metabolomics to compare the serum metabolite profile between young and adult CF (n = 39) patients and healthy controls (n = 30 ). Besides, we compared the metabolite profiles between the CF and CF Classes III and IV. Results: A distinctive metabolic profile between the groups was ascertained using univariate and volcano plot analyses that showed alterations in 79, 21, and 13 significantly differentially regulated metabolites between the patients with CF and healthy controls, within CF classes, and between CF Class III and IV, respectively. The identified metabolites included purines, amino acids, di, and tripeptides, involved in nitrogen oxide (NO), glutamine, glutamate, and arginine metabolism. The top metabolites include1-Aminopropan-2-ol, ophthalmate, serotonin, cystathionine, gamma-glutamylglutamic acid. Conclusion: The metabolomic profile between the patients with CF and healthy controls and between the functional classes has identified alterations of amino acids and dipeptides, involved in regulating glutathione, NO, and identified pathways altered in each state. Understanding these specific metabolomic changes will provide a better understanding of the disease and targets for drug therapy.

"Preserving the in vivo metabolome and energy-sensitive phosphoproteome requires rapid freezing of tissue samples" (15)


Adam J Rauckhorst, The University of Iowa (Primary Presenter)

Alora Kraus, University of Iowa

Diego Scerbo, University of Iowa

Eric Taylor, University of Iowa


Metabolomic regulation is rapid, highly dynamic, and intrinsically linked to oxygen as the terminal electron acceptor of the electron transport chain. Thus, loss of oxygen supply quickly propagates to broad metabolomic disruption. However, the timing of broad hypoxia-induced metabolomic changes elicited by tissue dissection are not well described. The implications of this absence for past and current metabolomic inquiries across the life sciences are far reaching. Here, we utilize GC- and LC-MS based metabolomics to examine the stability of the mouse liver metabolome during 10 minutes of acute hypoxia following tissue dissection. Striking metabolomic changes, detected by both principle component analysis and examination of individual metabolite levels, occurred within 30 seconds of tissue dissection. The magnitudes of these changes were amplified with time. Decreasing ATP levels coincided with markedly increasing AMP levels. Furthermore, glycolytic metabolites generally increased, whereas TCA cycle intermediate metabolites decreased, apart from succinate. Changes in the abundance of numerous additional metabolites are consistent with increased anaerobic metabolism resulting from continued ATP demand during inhibited oxidative phosphorylation with initial maintenance of ATP levels via adenylate kinase activity. The net accumulation of AMP resulted in activation of AMPK and phosphorylation of its downstream targets. Finally, AMP-fueled purine degradation metabolites were of the highest fold increased and may serve as a quantitative readout of sample handling and quality. These results demonstrate the metabolome is exquisitely sensitive to hypoxia caused by loss of perfusion from dissection. This study illustrates the importance of freezing samples within seconds of dissection to obtain the most accurate portrait of the in vivo metabolome.

"Metabolic Pathways Associated with Pain, Fatigue, Anxiety, and Depression in Children with Cancer" (17)


Ronald Eldridge, Emory University (Primary Presenter)

Jinbing Bai, Emory University

Canhua Xiao, Yale University

Sharon M. Castellino, Children's Healthcare of Atlanta

Janice Withycombe, Clemson University


Introduction: Over 16,000 children and adolescents are diagnosed with cancer yearly in the U.S. The majority are cancers of the blood or solid tumors that require chemotherapy for systemic treatment. Children who undergo chemotherapy can suffer moderate to severe psychoneurological symptom toxicities such as pain, fatigue, anxiety, and depression which may have lasting effects on survivors as they age into adulthood. The purpose of this analysis is to study the metabolic underpinnings related to these four symptoms in children with cancer.

Methods: Forty children with different cancer types participating in a larger study assessing symptoms and active cancer therapy at Children’s Healthcare of Atlanta were recruited for this sub-study. Pain, fatigue, anxiety, and depression were assessed using the Pediatric PROMIS measures (i.e., T-score metric with mean=50 and standard deviation=10). As these symptoms often occur together, a cluster variable was also created: “yes” for reporting T-score>50 for any one symptom, otherwise, “no”. High-resolution untargeted metabolomics was performed on serum collected <6 weeks of starting chemotherapy treatment according to the HILIC-positive chromatography high-performance mass spectrometry protocols of the Jones Clinical Biomarkers Laboratory at Emory University. Metabolite feature peak extraction and quantification was performed by adaptive processing software (apLCMS). The relative intensities of the metabolites were log2 transformed and quantile normalized. We used multivariable linear regression adjusted for age and sex to assess metabolites associated with the four individual symptoms; logistic regression was used for the binary symptoms cluster variable. Features that met a P-value<0.05 threshold were used to identify disrupted metabolic pathways using Mummichog network algorithms; we report pathways that met a P<0.05 significance. Given the discovery nature of this project, we did not correct for multiple comparisons.

Results: The children ranged from 7-18 years of age (mean=13.2 years). Twenty-two were girls (55%). A total of 9,276 unique m/z-retention time features were extracted from the samples and analyzed. Three hundred twenty-nine metabolites were associated with pain, 426 with fatigue, 558 with anxiety, 953 with depression, and 259 with the symptoms cluster variable. The most significant disrupted metabolic pathways were related to fatty acids. These pathways were associated with pain but not the other symptoms: fatty acid activation (P=0.006), fatty acid metabolism (P=0.009), and fatty acid biosynthesis (P=0.015). Amino acid metabolic pathways were disrupted in multiple symptoms. Tryptophan and methionine/cysteine metabolism were associated with the cluster variable (P=0.034 and P=0.049, respectively). Purine metabolism was associated with pain (P=0.021). Valine/leucine metabolism was associated with anxiety and depression (P=0.012 and P=0.055, respectively).

Conclusions: This preliminary analysis suggests that pediatric patient reports for pain, anxiety, and depression may be, in part, linked with key fatty acid and amino acid metabolic pathways. The fatty acid pathways involved metabolites of linoleic acid, an essential omega-6 fatty acid whose metabolism has been previously linked to inflammation and pain in humans. Some of the disrupted amino acid pathways found in this analysis previously have been associated with psychoneurological symptoms via the microbiome gut-brain axis. The discovery nature of these findings require additional investigation.

"Serum Metabolic Signatures of Chronic Limb-Threatening Ischemia in Patients with Peripheral Artery Disease" (19)


Philip Britz-McKibbin, McMaster University Hamitlon, Canada


Peripheral artery disease (PAD) is characterized by atherosclerotic narrowing of lower limb vessels leading to ischemic muscle pain in older persons. Some patients experience progression to advanced chronic limb-threatening ischemia (CLTI) with poor long-term survivorship. Herein, we performed serum metabolomics to reveal the mechanisms of PAD pathophysiology that may improve its diagnosis and prognosis to CLTI complementary to ankle-brachial index (ABI) and clinical presentations. Non-targeted metabolite profiling of serum was performed by multisegment injection-capillary electrophoresis-mass spectrometry (MSI-CE-MS) from age and sex-matched, non-diabetic PAD participants who were recruited and clinically stratified based on the Rutherford classification into CLTI (n=18) and intermittent claudication (IC, n=20). Compared to non-PAD controls (n=20), PAD patients had lower serum concentrations of creatine, histidine, lysine, oxoproline, monomethylarginine, as well as higher circulating phenylacetylglutamine (p < 0.05). Importantly, CLTI cases exhibited higher serum concentrations of carnitine, creatinine, cystine and trimethylamine-N-oxide along with lower circulating fatty acids relative to well-matched IC patients. Most serum metabolites associated with PAD progression were also correlated with ABI (r = ± 0.24-0.59, p < 0.05), whereas the ratio of stearic acid to carnitine, and arginine to propionylcarnitine differentiated CLTI from IC with good accuracy (AUC = 0.87, p = 4.0 × 10-5). This work provides new biochemical insights into PAD progression for early detection and surveillance of high-risk patients who may require peripheral vascular intervention to prevent amputation and premature death.

"Cysteine catabolism and the serine biosynthesis pathway support pyruvate production during pyruvate kinase knockdown in pancreatic cancer cells" (45)


Elliot M Ensink, Michigan State University (Primary Presenter)

Lei Yu, Michigan State University

Shao Thing Teoh, Michigan State University

Martin Ogrodzinkski, Michigan State University

Che Yang, Michigan State University

Ana Vazquez, Michigan State University

Sophia Lunt, Michigan State University


Pyruvate kinase, especially PKM2, is highly expressed in PDAC cells, but its role in pancreatic cancer remains controversial. To investigate the role of pyruvate kinase in pancreatic cancer, we knocked down PKM2 individually as well as both PKM1 and PKM2 concurrently (PKM1/2) in cell lines derived from a Kras G12D/-; p53 -/- pancreatic mouse model. We found that PDAC cells are able to proliferate despite PKM2 and PKM1/2 knockdown. Using liquid chromatography tandem mass spectrometry, we further uncovered that PKM1/2 knockdown cells continue to produce pyruvate. We explored the contributions of alternative pathways to pyruvate production, such as the serine biosynthesis pathway. Knockout of phosphoglycerate dehydrogenase in the serine biosynthesis pathway decreased pyruvate production from glucose, suggesting that the serine biosynthesis pathway contributes to pyruvate production from glucose during PKM1/2 knockdown. We further discovered that cysteine degradation generates ~20% of intracellular pyruvate. These results highlight the ability of cancer cells to adaptively rewire their metabolic pathways during knockdown of a key metabolic enzyme.

"Exploring metabolic differences in varying grades of meningiomas through nontargeted mass spectrometry" (69)


Zach Rabow, UC Davis (Primary Presenter)

Mirna Lechpammer, New York University Langone Medical Center

Oliver Fiehn, UC Davis\


Rationale: Emerging molecular data demonstrates the importance of genomic and epigenetic factors in pathogenesis of meningioma. Understanding the metabolic landscape using mass spectrometry is needed to overcome significant uncertainty in predicting tumor behavior and risk of recurrence. Current technologies lack sensitivity and selectivity, which hinders discovery of potential novel diagnostic and prognostic features. Here, we present a nontargeted metabolomic approach applied to meningiomas for biomarker discovery and identification of potential therapeutic targets.

Methods: Fresh frozen tissue from 36 patients (57% women, 43% men; mean age: 48) who underwent surgical resection for newly diagnosed meningiomas and 9 patient samples with non-neoplastic dura were used for case/control comparison. Additionally, 16 patient derived meningioma cell lines (Grade I-III), 2 immortalized human cell lines (IOMM-Lee and Ch157MN), and immortalized arachnoid cells were used to further explore metabolic changes in meningiomas. Metabolites were extracted from tissue and cells for GC-TOF (primary metabolism), RPLC ESI (±) (lipidomics) and HILIC ESI (±) (biogenic amines) coupled to high resolution mass spectrometry for analysis. Metabolites were identified using authentic standards, retention time, and MS2 fragmentation.

Results: Over one thousand known metabolites were identified and annotated as well as over 300 unknown metabolites. Metabolites were grouped into one of fifteen classes based on chemical ontology and function. Bis(monoacylglycero)phosphates were over 2-fold increase in atypical (Grade II) meningiomas versus Grade I, indicating lysosomal activation. One carbon metabolism pathway showed significant upregulation in neoplastic tissue vs. tumor involved dura, as well as neoplastic cell lines compared to arachnoid cells, indicating folate-dependent pathways as a potential therapeutic target.

Conclusions: Using novel combined untargeted metabolomics, we found multiple classes of metabolites that were either enriched or suppressed in neoplastic tissue and cells compared to non-neoplastic cells. Further studies are warranted for better understanding of possible oncogenic signaling pathways and to detect potential biomarkers useful for diagnosis and treatment.

"Annotation of serum steroid and neutral lipid metabolites from the Isle of Wight multigenerational birth cohort" (75)


Thilani Anthony, Department Of Biochemistry and Molecular Biology, Michigan State University (Primary Presenter)

Wilfried Karmaus, Division of Epidemiology, Biostatistics, and Environmental Health, School of Public Health, University of Memphis, Memphis, Tennessee, United States

Su Chen , Department of Mathematical Sciences, University of Memphis, Memphis, Tennessee, United States

Susan Ewart , Department of Large Animal Clinical Sciences, College of Veterinary Medicine, Michigan State University, East Lansing, Michigan 48824, United States

Syed Hasan Arshad , 1. Clinical and Experimental Sciences, Faculty of Medicine, University of Southampton, Southampton, Hampshire, United Kingdom 2. The David Hide Asthma and Allergy Research Centre, Isle of Wight, Hampshire, United Kingdom 3. NIHR Respiratory Biomedical Research Unit, University Hospital Southampton, Southampton, Hampshire, United Kingdom

John Holloway , Human Development and Health, University of Southampton, Southampton, Hampshire, United Kingdom

Hongmei Zhang, Division of Epidemiology, Biostatistics, and Environmental Health, School of Public Health, University of Memphis, Memphis, Tennessee, United States

Arthur Daniel Jones, Michigan State University


Prenatal exposures to metabolites, nutrients, and toxins (MNTs) have been linked to changes in the fetus that alter its epigenetic make-up, increasing risks of developing asthma in children and chronic obstructive pulmonary disease (COPD) in adults. One potential mechanism involves exposure-driven addition of a methyl group to cytosine-phosphate-guanine (CpG) sites in DNA that result in changing their gene activity or alternate splicing. Although numerous exogenous and endogenous MNTs are anticipated, it remains uncertain which in utero MNT exposures are associated with differential DNA methylation at birth and lung function later in life. Herein, we describe mass spectrometry-based approaches to measure exogenous and endogenous MNTs in maternal and cord blood specimens in parallel with measurement of DNA methylation and lung function in offspring. Serum specimens from two generations (maternal sera from F0 grandparents at the time of birth of their F1 offspring and cord sera collected at birth of F2 children) were provided by the Isle of Wight population-based birth cohort (IOWBC) study. This study provides a unique opportunity to exploit serum profiles collected nearly 30-years apart. MNTs were extracted from sera using a modified Matyash two-phase protocol into organic- and water-soluble components. Blanks, reference serum, and pooled QC sera were analyzed with each batch. Organic-soluble fractions were analyzed using LC-MSE (Waters G2-XS QToF) using a C18 LC column in negative-ion mode and separately using flow injection analysis in positive-ion mode. Peak alignment, detection, normalization, and annotation were performed using Progenesis QI (Waters) software. LC/MS profiling of serum MNTs in negative-ion mode revealed a wide assortment of endogenous lipids (phospholipids, lyso-phospholipids, free fatty acids, sphingomyelins, ceramides), steroid sulfates and glucuronides, bile acids, amino acids, drug metabolites, and products of microbial metabolism. Untargeted analysis yielded information about levels of 18 detected steroids, primarily sulfate and glucuronide conjugates of cholesterol and an assortment of downstream metabolites. Since many neutral lipids (triglycerides, cholesteryl esters and neutral steroids) do not ionize in negative-ion mode, positive-ion analysis using flow injection analysis provided information about levels of 62 annotated signals, mainly triglycerides and cholesteryl esters. Comparisons of levels in the F0 and F2 sera revealed systematic differences suggestive of metabolite degradation during long-term storage. These findings suggest alternatives for MNT normalization may be warranted for long-term longitudinal MNT profiling and recommend improved approaches to stabilizing specimen compositions for long-term longitudinal investigations.

"Metabolic Signatures are Associated with Persistent Critical Illness in Patients with Sepsis" (83)


Theodore Salvatore Jennaro, University of Michigan College of Pharmacy (Primary Presenter)

Elizabeth M. Viglianti, University of Michigan

Nicholas E. Ingraham, University of Minnesota - School of Medicine

Michael A Puskarich, University of Minnesota - School of Medicine

Alan E. Jones, University of Mississippi Medical Center

Kathleen A. Stringer, University of Michigan


Background: Persistent critical illness (PerCI) occurs when a patient’s continued dependency on the intensive care unit (ICU) becomes less related to their admitting diagnosis and more associated with an ongoing cascade of new onset organ failures. At the population level, this occurs during the second week of ICU admission. Such prolonged ICU stays cause significant patient morbidity and place a financial burden on the health care system. Patient characteristics such as age, comorbidities, and illness severity on ICU admission have not consistently identified who will develop PerCI. The application of metabolomics to PerCI may help meet this unmet critical need and provide insight into the physiological mechanisms underlying patient heterogeneity.

Methods: We conducted a metabolomics analysis of a multicenter clinical trial that randomized patients with septic shock to low (6g), medium (12g), and high dose (18g) Levo-carnitine or saline placebo. Serum samples were collected at baseline and assayed for acylcarnitines by liquid-chromatography mass spectrometry (LC-MS) and energy metabolites and amino acids using quantitative 1H-nuclear magnetic resonance (NMR). Development of PerCI was defined as an ICU stay of at least 10 days. Patients not developing PerCI were stratified by mortality status: Alive and discharged from the ICU prior to the development of PerCI or Died prior to the development of PerCI. A one-way analysis of variance (ANOVA) for each metabolite measured determined differences in concentrations stratified by PerCI outcomes. Post-hoc testing was done according to Fisher and p-values were corrected for multiple comparison according to the false discovery rate (FDR) procedure of Benjamini–Hochberg. We then conducted multivariable multinomial logistic regression to determine the effect of metabolite signatures after adjusting for clinical covariates. Covariates included illness severity (as measured by the Sequential Organ Failure Assessment Score) and a modified version of the Charlson Comorbidity Index.

Results: Of 247 patients, 90 (36.4%) developed PerCI, while 85 (34.4%) were discharged alive and 72 (29.2%) died prior to the development of PerCI. Patient characteristics including dose of L-carnitine received, gender, BMI, and ethnicity were similar among the outcome groups. Metabolomics data were available on a subset of patients (N=236 for acylcarnitines and N=228 for 1H-NMR). For acylcarnitine and NMR assay, 3/24 (12.5%) and 2/27 metabolites (7.4%) were different among all three outcome groups (FDR<0.05). These differentiating metabolic signatures were propylene glycol and short chain acylcarnitines including acetylcarnitine (C2), propionylcarnitine (C3), and butyrylcarnitine (C4). In multivariable modeling, baseline acetylcarnitine concentration was associated with being discharged alive (Relative Risk Ratio, RRR = 0.610, 95% CI: 0.417-0.893) and dying (RRR = 1.377, 95% CI 0.971-1.954) relative to the development of PerCI. Similar trends were observed for the other metabolites.

Conclusion: Alterations in short chain acylcarnitines and propylene glycol were associated with the development of PerCI and early death in patients with sepsis, which indicates altered β-oxidation of fatty acids and deranged energy metabolism contribute to heterogenous outcomes in the ICU. Our work highlights that metabolomics may be useful in identifying patients at risk of developing PerCI and the pathways contributing to disease pathophysiology.

"Innovations and Applications for Metabolomics and Exposomics Studies" (87)


Jennifer Reid, The Metabolomics Innovation Centre


Background: The Metabolomics Innovation Centre (TMIC) is Canada’s national metabolomics platform. TMIC is housed in 7 nodes at 4 universities, including University of Alberta, University of Victoria, McGill University and McMaster University. TMIC provides metabolomics services to academic and industry researchers and develops novel technologies for metabolomic profiling. TMIC also develops and maintains a wide range of freely available metabolomics databases and web services, including HMDB, DrugBank, T3DB and MetaboAnalyst.

Objective: TMIC has developed high throughput, low cost quantitative metabolomics assays and kits for a wide range of applications including human health, epigenetics, veterinary, agriculture and livestock. Our objective is to streamline metabolomics and provide all components required except instrumentation to enable metabolomics research and facilitate the translation of research discoveries. Here we will describe our LC-MS, NMR, and GC-MS assays and kits recently developed at TMIC.

Methods: TMIC’s LC-MS/MS kit contains all the components required for MS analysis, including standards, internal standards, QCs, MS running solution and detailed instructions. The kit quantitatively measures up to 154 metabolites. This kit is compatible with serum, plasma, and urine samples. TMIC’s NMR kit contains all the components required for 1 H NMR analysis, including buffer solution, deuterated internal standards and detailed instructions. The kit also contains an access code for a web-based, automated NMR spectral profiling server. It automatically determines the concentration of NMR-detectable metabolites accurately (~95% correct identification; ~10% quantification error) in < 3 minutes per sample. This NMR kit is compatible with serum, plasma, and cerebrospinal fluid samples. TMIC’s GC-MS kit includes a derivatization reagent, internal and alkane standards (C8-C20 and C22-C40), and detailed instructions. The kit contains an access code for GC-AutoFit, an automated, web-based tool, which can identify and quantify up to 120 compounds in urine, serum, plasma, CSF and milk.

Summary: TMIC’s quantitative metabolomics kits provide the ability to analyze over 250 metabolites using 3 different analytical platforms, including biologically relevant metabolites such as amino acids, acylcarnitines, phospholipids, biogenic amines, TCA cycle intermediates, sugars, organic acids and markers of exposures. These kits are available through TMIC for academic and industry research use. We expect these kits will enable a wealth of new metabolomics applications in both research and clinical settings.

"Towards Cancer Comprehensive and Early Diagnoses with NMR-based Metabolomics" (105)


Leo L Cheng, Massachusetts General Hospital

Charlestown, United States


Last year, at the 1st MANA annual meeting, we reported our NMR-based metabolomic results on human prostate (PCa) and lung (LuCa) cancers in the past 15+ years. Our aim of human cancer metabolomic studies has been the attempt to develop metabolomics in assisting the personalized precision medicine by providing disease characterizations at pre-symptomatic stages. Results towards this aim for human prostate and lung cancers will be presented.

Detecting Clinically Significant PCa: Tissue Metabolomics Refines Multiparametric Magnetic Resonance Imaging (mpMRI)-Ultrasound (US) Fusion Prostate Biopsy (Bx). mpMRI-US fusion prostate Bx has increased detection of clinically significant PCa, in part by revealing tissue structure and water property changes potentially associated with PCa. However, this approach does not assess disease biomolecular activities. Using clinical, mpMRI-positive, fusion Bx-targeted tissue cores and mpMRI-negative controls from 66 consecutive cases, we studied the potential of intact tissue magnetic resonance spectroscopy (MRS) to yield cancer metabolomic profiles that could help discriminate likely indolent from clinically significant PCa. Our MRS-based PCa metabolomic analyses, performed prior to histology, were able to: (1) estimate the scale of PCa metabolomic fields, important for future implementations of in vivo imaging, and (2) significantly differentiate clinically significant disease, based on patient follow up data to March 2020, in tissues deemed benign or low-risk PCa by pathology and imaging. These findings support the use of MRS-based metabolomic analysis to refine clinical PCa characterization, better discriminate between benign and aggressive tumors, and thereby assist patients and physicians in developing personalized surveillance or aggressive treatment plans.

Investigation of Human LuCa NMR-based Serum Metabolomics Prior to Disease Diagnosis. LuCa is currently the leading cause cancer deaths. Early stages are mostly pre-symptomatic and contribute to a delayed diagnosis leading to 70% of the patients dying from LuCa. A simple blood test provides good features as a screening method with being minimally-invasive, cost effective and without radiation exposure. Our previous results showing that metabolomic profiles could better differentiate LuCa from controls and between different cancer type encouraged us to evaluate pre-symptomatic patient serum samples prior to LuCa diagnosis. Twenty-five paired LuCa serum samples from patients prior to (PriorDx) and at time of (AtDx) a LuCa diagnosis were measured with matched Healthy controls. High resolution magic angle (HRMAS) 1H MRS serum spectra were measured. Spectral regions with >80% samples presented ≠0 values were further analyzed. Spectra from AtDx and Healthy were used as the training cohort, and subjected to PCA and the most significant six PCs were continued with Canonical Analysis. Spectra from PriorDx samples formed the testing cohort and passively followed all the calculation steps of the training cohort. Canonical results shown that PriorDx resides between Healthy and AtDx, differing significantly from each with p values 0.022 and 0.00024, respectively , and overall accuracy represented by the ROC with AUC 0.69 and 0.87, respectively. By using M±SE, thus calculated, as a threshold, and patient follow up data to June 2020, survival estimates were calculated. These results proposed a testable hypothesis that centers serum MRS for use in LuCa early detection screening.

"Vitamin D and 17q21 genetic variances are associated with plasma sphingosine-1-phosphate and ceramide concentrations" (133)


Mengna Huang, Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital (Primary Presenter)

Antonio Checa, Department of Medical Biochemistry and Biophysics, Division of Physiological Chemistry II, Karolinska Institutet, Sweden

Pei Zhang, Department of Medical Biochemistry and Biophysics, Division of Physiological Chemistry II, Karolinska Institutet, Sweden; Gunma University Initiative for Advanced Research (GIAR), Gunma University, Japan

Rachel Sabine Kelly, Channing Division of Network Medicine, Brigham and Womens Hospital

Su Chu, Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA, USA

Priyadarshini Kachroo, Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA, USA

Bo Chawes, COPSAC, Copenhagen Prospective Studies on Asthma in Childhood, Herlev and Gentofte Hospital, University of Copenhagen, Copenhagen, Denmark

Klaus Bønnelykke, COPSAC, Copenhagen Prospective Studies on Asthma in Childhood, Herlev and Gentofte Hospital, University of Copenhagen, Copenhagen, Denmark

Augusto Litonjua, Division of Pediatric Pulmonary Medicine, Department of Pediatrics, Golisano Children's Hospital at Strong, University of Rochester Medical Center, Rochester, NY, USA

Scott Weiss, Channing Division of Network Medicine, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA, USA

Hans Bisgaard, COPSAC, Copenhagen Prospective Studies on Asthma in Childhood, Herlev and Gentofte Hospital, University of Copenhagen, Copenhagen, Denmark

Craig Wheelock, Department of Medical Biochemistry and Biophysics, Division of Physiological Chemistry II, Karolinska Institutet, Sweden; Gunma University Initiative for Advanced Research (GIAR), Gunma University, Japan

Jessica Lasky-Su, Brigham and Womens Hospital


Sphingolipids play important roles in influencing the risk of childhood asthma. Chromosome region 17q21 (ORMDL3) is a widely replicated asthma-susceptibility locus for childhood-onset asthma with several identified polymorphisms, and the encoded protein suppresses serine palmitoyl-CoA transferase (rate-limiting enzyme in de novo sphingolipid synthesis). Additionally, experimental studies have shown that vitamin D metabolites can activate the sphingolipid pathway, and 17q21 genotype may modify the effect of vitamin D in preventing childhood asthma/recurrent wheeze in early life. In this study, we aim to investigate associations between vitamin D and sphingolipid concentrations in plasma, and disentangle the impact of single nucleotide polymorphisms (SNPs) in the 17q21 region on these relationships.

Sphingolipid concentrations were measured in plasma samples from children participating in the Vitamin D Antenatal Asthma Reduction Trial (VDAART) at age 6 years, using a targeted liquid chromatography coupled tandem mass spectrometry platform quantified with internal standards. The standardized levels of 77 sphingolipids were associated with concurrent 25-hydroxyvitamin D [25(OH)D] (considered most reliable biomarker of vitamin D status) levels using linear regression models adjusted for child sex, race, ethnicity, study site, sample quality (color indication of hemolysis), and concurrent body mass index, with subgroup analyses conducted within categories of four pre-selected 17q21 SNPs. The same sphingolipids were also tested for associations with the SNPs using linear regression models adjusted for child sex, race, ethnicity, and sample quality. Replication analysis was conducted in the similarly-designed Copenhagen Prospective Studies on Asthma in Childhood 2010 (COPSAC2010), with targeted sphingolipids data from the same platform.

Data from 421 children in VDAART with targeted sphingolipid measurements were analyzed as the discovery population, including 366 with genetic data; 501 children from COPSAC2010 were included as the replication population (442 with genetic data). Account for multiple comparisons at a significance threshold (p-value<4.17x10-4) derived from an effective number of independent tests accounting for 80% variance in sphingolipids, two species of sphingosine-1-phosphate (S1P) and four species of ceramides were associated with plasma 25(OH)D levels. Among these, with a replication threshold of p-value<0.05, S1P (d16:1) was positively associated with 25(OH)D consistently in VDAART (p-value=3.64x10-4) and COPSAC2010 (p-value=8.27x10-3), while ceramide (d18:2/16:0) had inverse associations with 25(OH)D in both cohorts (p-value=3.62x10-3 in VDAART, p-value=2.51x10-4 in COPSAC2010). Additionally, three ceramides and one dihydroceramide were negatively associated with 25(OH)D at nominal significance in both cohorts: ceramide (d18:1/20:0), ceramide (d18:1/22:0), ceramide (d18:1/24:0), and dihydroceramide (d18:0/16:0). In general, the negative associations between 25(OH)D and plasma ceramides levels were only observed in children with at least one copy of the high-asthma-risk C allele for rs12936231 and rs12603332 within the 17q21 region. Both SNPs by themselves were positively associated with ceramide (d16:1/24:1), ceramide (d18:1/22:0), and ceramide (d18:1/26:0) at nominal significance in VDAART.

Consistent with previous experimental evidence, plasma 25(OH)D was positively associated with S1Ps, and negatively associated with multiple ceramides. Given the complexity of the sphingolipid synthesis and recycling pathways, our analyses provided novel insight into how vitamin D and 17q21 genotype may influence plasma levels of different species of sphingolipids, which may in turn alter the risk of childhood asthma.

"Untargeted Metabolomics Analysis of First-trimester Serum to Discover Biomarkers and Mechanism of Pregnancy Complications: A Case-Control Study" (137)


Wimal Pathmasiri, Nutrition Research Institute, University of North Carolina at Chapel Hill, Kannapolis, NC, (Primary Presenter)

Yuanyuan Li, University of North Carolina at Chapel Hill

Emily Harville, Department of Epidemiology, Tulane University School of Public Health and Tropical Medicine, New Orleans, LA

Ke Pan, Department of Epidemiology, Tulane University School of Public Health and Tropical Medicine, New Orleans, LA

Susan Mcritchie, Nutrition Research Institute, University of North Carolina at Chapel Hill, Kannapolis, NC,

Susan Sumner , UNC Chapel Hill


Hypertensive disorders of pregnancy (HDP) and preterm birth (PTB) are two major obstetric complications with high morbidity and mortality. However, predictive biomarkers for early diagnosis as well as the underlying biological mechanisms are not known. This study aims to identify a) potential biomarkers from first-trimester serum to predict the complications; b) possible life exposures that are associated with risks; and c) perturbation of metabolic pathways. First-trimester serum obtained from 104 cases (51 HDP, 53 PTB) and 109 control subjects were selected from a biorepository. Untargeted liquid chromatography coupled high-resolution-Orbitrap-mass-spectroscopy (LC-HRMS) and nuclear magnetic resonance (NMR) were used to conduct profiling for both endogenous and exogenous metabolites. Univariate and multivariable logistic regression were used to identify metabolites that differed between phenotypic groups. Over 20 LC-HRMS peaks were deemed to be highly predictive for HDP and PB. Among the exogenous exposures, phthalate and salicylamide were highly related to the risk of HDP; while (R,S)-N-Acetyl-S-(2-hydroxy-3-buten-1-yl)-L-cysteine was related to PTB, although none significantly added to the predictive value of the models. Pathway analysis indicated that HDP and PTB shared common perturbations in pathways of protein biosynthesis, amino acid metabolism, and electrolyte equilibrium. HDP was highly related to perturbations in cholesterol and lipid metabolism, while PB was associated with inflammation and immune response.

"Response of the Plasma and Skeletal Muscle Metabolome to Exercise and Cardiorespiratory Fitness" (155)


Charles R Evans, University of Michigan (Primary Presenter)

Heidi IglayReger, University of Michigan

Jennifer LaBarre, Department of Nutritional Sciences, University of Michigan School of Public Health, Ann Arbor MI

Johanna Fleischman, University of Michigan

Charles Burant, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor MI


Regular exercise and physical fitness are associated with multiple health benefits, including reduced risk for cardiovascular disease, prevention or amelioration of diabetes, improved mental health, and many other factors. However, neither the molecular response to exercise nor the mechanisms by which it brings about these health benefits have been fully characterized. The metabolome is profoundly responsive to exercise and may serve as a mediator of some of its effects. Generating a more detailed view of the metabolome during exercise is therefore an important research objective.

Purpose. The purpose of this study was to assess the response of the human metabolome to a bout of high-intensity exercise, and to investigate differences between the metabolome of individuals with higher and lower cardiorespiratory fitness (CRF) as measured by VO2max.

Methods. We recruited 50 young healthy males, ages 18-30 and measured their VO2max using indirect calorimetry during cycle ergometry exercise. Those with the highest (10) and lowest (9) CRF underwent detailed phenotyping during cycle ergometry exercise including blood sampling every 3 minutes and vastus lateralis biopsy before and after exercise. Untargeted reversed phase and HILIC reversed-phase LC-MS based metabolomics were used to perform relative quantitative analysis of metabolites. Statistical analysis and curve-shape modeling were used to categorize metabolites that responded to acute exercise and/or differentiated low and high CRF groups.

Results: We generated a profile of the response of the plasma metabolome to exercise with high temporal resolution containing over 650 exercise-responsive features. Exercise-responsive metabolite classes included TCA cycle intermediates, amino acids, purines, nucleosides, and many unidentified features. Curve shape modeling differentiated these metabolites into distinct time- and fitness-dependent groups. A distinct subset of metabolites differentiated low and high CRF groups at rest.

Conclusion: Our study reveals the extensive response of the metabolome to physical activity and provides insight into the impact of cardiorespiratory fitness on the metabolome. Our findings contribute to the goal of uncovering molecular mediators of the health benefits of exercise.

“Understanding the impact of neurodegeneration on metabolism in Familial dysautonomia” (157)


Annie Waldum, Montana State University (Primary Presenter)

Alexandra Cheney, Montana State University

Stephanann Costello, Montana State University

Nick Pinkham, Montana State University

Seth Walk, Montana State University

Frances Lefcort, Montana State University

Valerie Copie, Montana State University


Crosstalks between cellular networks intersecting the gut-brain-liver axis have emerged as critical elements of neurodegenerative diseases, and new insights into the complexity of this axis could unravel routes for potential clinical interventions. Of great interest is the incurable neurodegenerative disease Familial dysautonomia (FD) that originates from a point mutation in the IKBKAP/ELP1 gene leading to tissue-specific reductions of the ELP1 protein, and affecting both the peripheral and central nervous systems. Prominent features of the disease, observed in both patients and FD mouse models, include major impairments of central metabolism and severe gastrointestinal dysmotility, which are supported by our findings that FD mice and patients struggle to maintain a normal weight and display severely altered gut microbiomes. We postulate that neuronal innervation deficits modulated by ELP1-neurodegeneation initiates a feedback loop whereby the gut microbiome becomes dysbiotic, contributing to impaired metabolism and an increase in neuronal mitochondrial stress, which further promote neurodegeneration. To better understand the molecular networks regulating the gut microbiome, liver metabolism, and neuronal health, a multidisciplinary approach has been undertaken to identify and interrogate the molecular mechanisms at the source of the neuropathy observed in FD. We have conducted a detailed 1H NMR metabolomics analysis of liver samples obtained from FD mouse models comparing them to the profiles of healthy, age-matched control mice. These experiments have identified significant changes in the liver metabolomes of the FD mice, which provide valuable insights into metabolic dysregulation and microbiome dysbiosis occurring in FD. Here, we present our latest metabolomics findings including a presentation of specific metabolic pathways found to be altered in our FD mouse models. Many neurodegenerative diseases share a range of molecular and cellular pathologies; our results have the potential to help unravel the dysregulations taking place between energy metabolism, gut microbiome and neurodegeneration in other more common neurodegenerative diseases such as amyotrophic lateral sclerosis, Alzheimer’s and Parkinson’s disease. Going forward, these findings may contribute to the development of novel therapeutics including clinical metabolic and/or microbiome interventions that delay or prevent neurodegeneration and comorbidities.

Funding: This research is supported by NIH grant R01 DK117473-01A1

Computational Approaches

"MetabolonR: a reproducible Jupyter-notebook workflow for analyzing metabolomics data" (33)


Fadhl Alakwaa, University of Michigan (Primary Presenter)

Sean Eddy, University of Michigan

Viji Nair, University of Michigan

Masha Georges Savelieff, University of Michigan

Stephen Goutman, University of Michigan

Brian Callaghan, University of Michigan

Matthias Kretzler, University of Michigan



Metabolomics is the large-scale study of metabolites in biofluids, tissue extracts, or organisms, and provides valuable insight and a deeper understanding of pathomechanisms underlying complex diseases, such as diabetes and chronic kidney disease. There has been a recent increase in the number of publications using metabolomics data for different applications, including precision medicine, biomarker discovery, and drug development and discovery. However, reproducibility of published results is a major challenge in metabolic research, owing to a lack of widely used standardized public data and code repositories used for metabolomic research that are common in other -omics fields. Here, we propose a standardized, open source analysis workflow that can be used as a template for any exploratory metabolomics data analysis. This pipeline performs quality checks and preprocessing steps, biomarker discovery using four methods, source of variation analysis, clustering and machine learning classification, and network analysis. It is implemented in Jupyter Notebook-R kernel and can be modified or extended for any application, and used to generate publication-quality figures. To demonstrate utility of the pipeline, we tested it using metabolomics data generated by Metabolon Inc. The samples were composed of 43 lean control versus 88 obese participants to identify metabolites that might be associated with diabetic neuropathy. The pipeline identified 5 metabolites that were associated with diabetic neuropathy including: 1,7-dimethylurate, caffeine, ergothioneine, glycocholate glucuronide (1), and theophylline.

Availability and Implementation: metabolonR is available at Github repository

"First principle quantum chemical modelling of electron ionization mass spectra for silylated compounds in GC-MS" (73)


Shunyang Wang, UC Davis (Primary Presenter)

Tobias Kind, UC Davis Genome Center - Metabolomics

Oliver Fiehn, UC Davis


Chemical derivatization, especially sialylation, is widely used in gas chromatography coupled to mass spectrometry (GC/MS). By introducing the trimethylsilyl (TMS) group to substitute active hydrogens in the molecule, thermostable and volatile compounds are created that can be easily analyzed. While large GC-MS libraries are available, the number of TMS-derivatized spectra is comparatively small. In addition, many metabolites cannot be purchased to produce authentic library spectra. Therefore, computationally generated in-silico mass spectral databases need to take TMS derivatizations into account for metabolomics. The quantum chemistry method (QCEIMS) is an automatic method to generate electron ionization (EI) mass spectra directly from compound structures. Here, we have built an in-silico EI-MS database of trimethylsilyl derivatives that include various compound classes such as acids, alcohols, amides, amines and thiols. We used additional subclasses according to the aromaticity and hybridization property (primary, secondary, tertiary and aromatic substitutions). We utilized mass spectral similarity scores and true positive rates to evaluate the prediction accuracies. We discuss fragmentation pathways found in the simulations according to the compound classes. We also introduce and test a new conformer sampling method combined with QCEIMS to provide an improved parallel multi-core calculation. The coverage of wide conformational spaces showed a limited effect on the final prediction accuracies. We discuss improvements such as the inclusion of excited state molecular dynamics using configuration interaction methods (CI) in the mass spectral predictions. While we show that QCEIMS can be a useful assistant tool to identify unknown compounds in GC/MS based metabolomics, in-silico spectral prediction accuracies must be further improved.

"A systematic approach for inferring absolute concentrations from relative abundances in metabolomics data" (97)


Justin Lee, Georgia Institute of Technology (Primary Presenter)

Mark Styczynski , Georgia Institute of Technology


As a direct readout of metabolic information, metabolomics has great potential to be used with metabolic modeling and engineering tools. Surprisingly, untargeted metabolomics has been sparingly used for these purposes. Perhaps the biggest roadblock is that the analytical methods used to capture untargeted metabolomics data, such as gas chromatography-mass spectrometry, are generally only able to measure relative abundances of metabolites and not absolute concentrations without significant experimental effort that is infeasible at the metabolome scale. While abundance fold changes of individual metabolites are useful for analysis and downstream interpretation, there is little, if any, quantifiable meaning when comparing relative abundances of different metabolites to each other. This can make it difficult to incorporate untargeted metabolomics data with current metabolic tools. While it is common to use standards in targeted metabolomics to measure some absolute concentrations, their use is limited by cost, commercial availability, and the time it takes to run each standard. A method for inferring absolute concentrations of metabolites without the need for standards would be incredibly valuable to the metabolomics community and for future metabolic modeling and engineering efforts. Here, we present our work in developing a systematic approach for calculating absolute quantities of intracellular metabolites from their relative abundance measurements. By using only metabolomics time course data and the stoichiometry of the system, we believe we can leverage mass balances and the dynamical nature of the data to determine the most likely absolute concentration profiles. Our preliminary results show that this approach can work on in silico data generated from small toy models. Before our approach can be applied to real metabolomics data, there are several challenges that need to be addressed. Noise in the data can cause discrepancies when solving mass balances, and because most metabolic systems contain more reactions than metabolites (i.e. they are underdetermined) it can be difficult to infer a single optimal solution of absolute concentrations. This work discusses our efforts to overcome these challenges and our next steps toward applying our framework to metabolomics data acquired from real biological systems.

"Automated metabolomic profiling of 1D 1H NMR spectra with MagMet" (121)


Mark Berjanskii, University of Alberta (Primary Presenter)

Manoj Rout, University of Alberta

Brian Lee, University of Alberta

Honeya Shahin, University of Alberta

David Wishart, University of Alberta

Rupasri Mandal, University of Alberta

Pascal Mercier, University of Alberta

Rustem Shaykhutdinov, University of Alberta

Nazanin Assempour, University of Alberta

Ying Dong, University of Alberta

Mathew Johnson, University of Alberta

Matthias Lipfert, University of Alberta


NMR is a powerful and widely used metabolomics technique, but is limited by the requirement of highly qualified personnel to operate the equipment and to perform time-consuming manual spectral profiling. To overcome these challenges, our lab has focused on the development of an easy-to-use automated web server, called MagMet, that can quickly, accurately, and automatically process 1H 1D NMR spectra of biofluids and generate extensive compound lists and concentrations. MagMet is able to automatically process raw NMR data, match the resulting spectrum against a spectral library of reference compounds and then automatically determine their concentrations. MagMet allows users to visually assess the quality of the automated spectral deconvolution process and, if necessary, adjust the NMR profile manually. The data can be exported in CSV and JSON formats. MagMet accepts NMR spectra collected on 500, 600, and 700 MHz instruments (Bruker or Varian). Extensive testing has shown that MagMet can automatically find the concentration of NMR-detectable metabolites accurately (~95% correct identification and ~15% quantification error), in less than 20 minutes. MagMet has been extensively optimized and validated for a wide range of biofluids such as serum/plasma, fecal water, cell growth media as well as consumer products such as wine and beer. MagMet is able to identify and quantify up to 58 compounds in normal serum/plasma (up to 140 compounds in diseased samples), 80 compounds in fecal samples and ~100 compounds in wine/beer. MagMet has been recently optimized for the analysis of human serum to diagnose dozens of inborn errors of metabolism. The total number of reference spectra in MagMet’s libraries is now nearly 200 molecules. Unlike manual profiling, the MagMet pipeline does not require human intervention. Because of its fully automated workflow, MagMet can save a lab thousands of dollars and hundreds of hours of an employee's time. We anticipate that MagMet will enable a resurgence of NMR-based metabolomics and increased interest in the use of NMR in clinical settings. MagMet is accessible at .

"Automation of Molecular Ionization Site Predictions based on Gibb’s Free Energies" (123)


Jessica Bade, Pacific Northwest National Laboratory (Primary Presenter)

Sean Colby, Pacific Northwest National Laboratory

Ryan Renslow, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA

Thomas Metz, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA


Thorough and precise molecular identification of unknown metabolites can unlock new areas of the molecular universe and allow greater insight into complex biological and environmental systems than currently thought possible. Analytical approaches for measuring the metabolome, such as nuclear magnetic resonance spectroscopy, and hyphenated separation techniques coupled with mass spectrometry, such as liquid chromatography-ion mobility spectrometry-mass spectrometry (LC-IMS-MS), have risen to this challenge by yielding rich experimental data that can be identified by cross-reference with similar information for known standards in appropriate reference libraries. Confident, unambiguous identification of molecules in metabolomics studies, though, is often limited by the diversity of available data across chemical space, the unavailability of authentic reference standards, and the corresponding lack of comprehensiveness of standard reference libraries. The In Silico Chemical Library Engine (ISiCLE) addresses the aforementioned hindrances by providing a first-principles, cheminformatics pipeline that yields collisional cross section (CCS) values for any given molecule and without the need for training data. In this program, chemical identifiers undergo molecular dynamics simulations, quantum chemical transformations, and ion mobility calculations for the generation of predicted CCS values. Here, we present a new module for ISiCLE that focuses primarily on the sensitivity of chemical property predictions to ionization site and adduct ion location and the corresponding effect ionization site can have on molecular CCS calculations. An update to adduct ion creation methods is proposed concerning a transition from pKa and pKb led predictions to a Gibbs free energy (GFE) based determinacy of true ionization site location. A validation set of molecular ionization sites was assembled from literature with experimentally confirmed protonation locations using methods of collision induced fragmentation pattern matching, infrared photon dissociation analysis, and high field asymmetric waveform ion mobility spectrometry analysis. The confirmed protonation sites were then cross-referenced with their respective pKb predicted locations and GFE values for all potential adduct ions. Upon evaluation of the two methods, the lowest GFE value, as based upon density functional theory calculations, was found to predict the true ion location with 100% accuracy while pKa had less accuracy. Furthermore, ionization site location was found to have an effect on the corresponding calculated CCS values and may be relevant to reducing the current error (3.2%) in calculated versus experimental CCS. The impact of potential mis-selection of ionization site location on final molecular property prediction values signifies the importance of acknowledging distinct structural populations when interpreting experimental data, such as from LC-IMS-MS measurements, and attributing molecular characteristics.

"RUMP: A Container-based Reproducible and Scalable Untargeted Metabolomics Data Processing Pipeline" (125)


Xinsong Du, Health Outcomes & Biomedical Informatics Department, College of Medicine, University of Florida (Primary Presenter)

Ke Xu, Health Outcomes & Biomedical Informatics Department, College of Medicine, University of Florida

Manfio Luran, Department of Health Outcomes & Biomedical Informatics, College of Medicine, University of Florida

Alexander Kirpich, Department of Population Health Sciences, School of Public Health, Georgia State University

William Hogan, Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida

Timothy Garrett, Univ of FL-Pathology

Dominick J. Lemas, Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida


Background: Reproducibility of untargeted metabolomics data processing remains a significant challenge, in part, due to limitation in informatics tools. Nextflow is a pipeline development tool that supports containerization and high performance computing to improve the scalability and reproducibility of computational workflows.

Objective: The goal of our project is to develop an open-source untargeted metabolomic data processing workflow using Nextflow that is both highly reproducible and scalable.

Findings: RUMP is a container-based reproducible and scalable untargeted metabolomics data processing pipeline. RUMP can be executed on any UNIX-like systems with a specific focus on implementation within large-scale computing environments. We have developed RUMP to run parallelized common metabolomics software packages including MzMine for peak detection, CEU Mass Mediator for metabolite identification, and Mummichog for pathway analysis. Moreover, we used the SECIMTools package for statistical tests such as blank subtraction, ANOVA and student t-test, as well as providing users multiple visualization methods including principle component analysis, volcano plots, hierarchical clustering. To facilitate dynamic and transparent data processing, we included MultiQC reports to visualize the result of data processing and summarize metabolomics output. Furthermore, the relationship between the input file size and the use of corresponding computing resources is monitored within the pipeline. This provides useful suggestions for further computing resource allocation and efficient use.

Conclusion: RUMP is a container-based platform that has potential to facilitate high-throughput and scalable untargeted metabolomics data analysis with high levels of reproducibility and scalability.

"Exploration of molecular conformer selection for accurate in silico chemical property prediction" (143)


Felicity Nielson, Pacific Northwest National Laboratory (Primary Presenter)

Dennis Thomas, Pacific Northwest National Laboratory

Sean Colby, Pacific Northwest National Laboratory

Ryan Renslow, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA

Thomas Metz, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA


A conformer, or 3D structure of a molecule, can drastically alter structure-dependent molecular properties. With a growing presence of computationally derived metabolite reference libraries, subtle differences between metabolite conformers can lead to misidentification in platforms such as ion mobility spectrometry (IMS), infrared spectroscopy, and nuclear magnetic resonance spectroscopy, where two unique metabolites might be neighbors in the property space being measured. For example, subtle differences in ion structure, even the rotation of a methyl group, can lead to significantly different measurements of molecular collisional cross sections (CCS). CCS is a measure that depends on shape, size, and chemistry of a molecular ion and is reported as a cross sectional area (Å2). It can be determined using IMS, which separates ions based on the extent of their interactions with a buffer gas (usually N2 or He) as they travel under the influence of an electric field. CCS is highly sensitive to molecular shape, showing measurable differences even between conformers.

The structural dependence of chemical reactions and property predictions makes choosing correct conformers for computational simulations critical. But to date, there has not been a clear study on the most appropriate method for conformer selection in computational pipelines. In order to increase calculated property accuracy, which is critical for metabolomics, small molecule design, biological activity assays, and many other research fields, we need to gain a better understanding of the best approaches for conformer selection. Here, we employed Monte Carlo simulation and physics-based approaches to perform an in-depth analysis of conformer sampling trends in the context of CCS prediction for a set of 18 molecules. We tested the performance of various conformer selection techniques – Boltzmann weighting, lowest energy selection, random selection, energy thresholds, and RMSD-based similarity down-selection – on structures generated by molecular dynamics (MD). While different selection techniques on MD conformers had different effects on precision and accuracy, no single technique or combination of techniques was sufficient to obtain levels of accuracy within experimental error. We found this was due to the randomness of their generation, even when sampling 50,000 conformers. As a small test, we ran Density Functional Theory (DFT) geometry optimization on 25,000-50,000 conformers of three small molecules and found this significantly improved the precision and accuracy of the selection techniques. We present our findings on which selection methodology performed best and discuss the current necessity of quantum chemistry geometry optimization or more efficient conformer generation tools.

"Entryway to adaptable, task-specific metadata management for metabolomics" (147)


Abigail Moore, University of Georgia (Primary Presenter)

Amanda Shaver, University of Georgia

Brianna Morgan Garcia, Department of Chemistry, University of Georgia

GONCALO GOUVEIA, Department of Biochemistry and Molecular Biology, University of Georgia

Arthur S Edison, University of Georgia


Metabolomics experiments are susceptible to unwanted variation that can occur throughout an experiment from experimental design through data acquisition. To assess unwanted variation, researchers must document potential sources of variation through metadata, which describe the context in which an experiment was conducted. Current metadata management solutions pose barriers for researchers by requiring expertise to adapt a solution to task-specific needs, which could include collecting metadata that are unique to an experiment or relating metadata to efficiently retrieve the full lifecycle of a sample.

To mitigate barriers to metadata management in metabolomics, we developed a solution that combines the familiarity of spreadsheets with the integrity of a relational database. Our solution capitalizes on Google Sheets to collect metadata, which reduces the barrier of developing web interfaces for collaborative metadata collection. Backed by a relational database, our solution ensures the quality and integrity of metadata by enforcing consistent data types, suggesting ontology terms, and maintaining relationships among the metadata. Our implementation demonstrates metadata management from sample growth through data acquisition and considers minimum reporting standards and best practices established by the metabolomics community.

While there are many barriers to adapting a metadata management solution to the task-specific needs of individual experiments, our solution begins lowering those barriers. By these means, researchers have a greater opportunity to adapt a metadata management solution to their needs and ensure that their metadata are both fit for purpose, trustworthy and available to identify variation relevant to experimental design and data analysis. Finally, our solution promotes achieving metabolomics metadata standards and makes experimental data align with FAIR (findable, accessible, interoperable, reusable) principles.

Ecology and Environment

"Site-specific metabolic effects of acute naphthalene exposure revealed by gross lung microdissection in mice" (31)


Nathanial Stevens, UC Davis (Primary Presenter)

Patricia Edwards, UC Davis Center for Health and the Environment

Lisa Tran, UC Davis Center for Health and the Environment

Laura Van Winkle, UC Davis Center for Health and the Environment

Oliver Fiehn, UC Davis


Naphthalene is a ubiquitous environmental contaminant produced by combustion of fossil fuels and is also a primary constituent of both mainstream and side stream tobacco smoke. Naphthalene elicits site-specific toxicity in airway club cells through CYP450-mediated bioactivation into a reactive epoxide, resulting in depletion of glutathione and subsequent cytotoxicity upon acute exposure. While effects of naphthalene in mice have been extensively studied, few experiments have characterized the metabolic response of the lung to naphthalene exposure. Furthermore, most studies of lung metabolism analyze homogenized whole lung tissue, which prevents detecting the response of individual lung regions targeted by a specific toxicant. Therefore, the response of the airways and parenchyma to intraperitoneal injections of naphthalene was assessed 2, 6, and 24 hours post-injection through untargeted metabolomics of gross microdissected lung tissue in male and female mice. Microdissected airway and parenchyma sections were analyzed by LC/MS/MS assays for both lipids and hydrophilic metabolites. In total, 616 unique annotated compounds were identified across all platforms. Most significantly altered metabolites in response to treatment were present at 2 and 6 hours post-injection, with significant differences in metabolite profiles of female airways persisting at 6 hours compared to 2 hours in male mice. Several lysophosphatidylcholines (LPCs), dipeptides, and oligopeptides were all significantly reduced in both the airways and parenchyma of the mice. The greatest changes in each compound class were most pronounced at 6 hours in the female airways. Additionally, many species of triacyl glycerides (TGs) were substantially increased exclusively in airways 6 hours after naphthalene exposure. These changes may correspond to alterations in membrane permeability, inflammatory response, and structural remodeling that could modulate some of the effects of repeated naphthalene exposure. Overall, this study recapitulates previous findings of naphthalene studies in mice and demonstrates the importance of site-specific analysis of the lung to characterize changes in metabolism following toxicant exposure.

"Metabolic Strategies Adopted By Cuscuta campestris Is Aimed at Remodelling Primary and Secondary Metabolism in The Host Species Artemisia campestris to Facilitate the Parasitization Process" (43)


Biswapriya Biswavas Misra, Department of Internal Medicine, Section of Molecular Medicine, Wake Forest School of Medicine (Primary Presenter)

Fabrizio Araniti, Department AGRARIA, University Mediterranea of Reggio Calabria, Locality Feo di Vito, 89122, Reggio Calabria (RC) – Italy

Marco Landi, Department of Agriculture, Food and Environment, University of Pisa, Pisa, Italy – Italy



Cuscuta campestris is a holoparasitic species that parasitizes several wild and crop species and causes crop and economic losses. Among its hosts, Artemisia campestris subsp. variabilis, a medicinal plant species of pharmaceutical and nutraceutical interest, is one of its affected hosts. Thought some information are available on recognition and anatomical alterations induced in the hosts, no data are reported on molecular interactions of Cuscuta and A. campestris or its mechanisms of parasitism or changes induced in host primary and specialized metabolism upon parasitization.


In the present study, using untargeted gas chromatography-mass spectrometry (GC-MS) metabolomics approach, we investigated both (a) volatile organic compounds (VOCs) involved in host-parasite interaction using headspace solid-phase microextraction (SPME) technique, and (b) metabolic changes induced in primary metabolism (silylated derivatization of polar extracts), in addition to recording several physiological parameters of the host species in their natural ecosystems on A. campestris parasitized with Cuscuta (n=4) and non-parasitized plants (n=4).


Our GC-MS metabolomic analysis captured the differential relative abundances of 148 metabolites that indicated parasitization induced significant alterations in the amino acid metabolism, strongly increased production of osmoprotectants (GABA, proline, oxoproline and mannitol, maltotriitol, arabitol, galactinol, myo-inositol), which generally accumulate in plants and are antioxidative. Further, significantly altered carbohydrate levels, i.e., panose, melibiose, galactose, mannose, melezitose, trehalose, fucose, psicose, raffinose, and turanose were observed, indicating parasite induced changes in primary metabolism. Of specialized metabolism, significantly altered levels of shikimic acid, coumaric acid, coniferin, ferulate, quinate, and 4-hydroxybenzoic acid- are all defense compounds that confer protection against pathogens, free radical scavenging, and provide mechanical reinforcement during host invasion. Finally, a significantly increased accumulation of trehalose, a simple monosaccharide, in parasitized plants was observed, which possibly interferes with the cell wall polysaccharides of Cuscuta vine tips causing necrosis and negatively impacts parasitization.

Further, the analysis of VOCs by SPME demonstrated a significant reduction of several metabolites involved in defense and Cuscuta parasitization. In particular, a reduction in the total sesquiterpenoid (characteristic specialized metabolites in Artemisia genus) content was observed in the host, indicating a suppression of host defense. On the contrary, the monoterpenoids that are involved in host recognition, were weakly affected by the parasitization. However, the monoterpene pinene’s production, a monoterpene involved in Cuscuta chemotaxis, was significantly increased in parasitized plants. Finally, invasion by Cuscuta inhibited the 3-hexen-1-ol acetate production in the host, a secondary metabolite with a known repellent activity against Cuscuta.

At the physiological levels, Cuscuta parasitization altered the photosynthetic machinery of the host that induced alterations of several parameters related to the photosystem II (PSII). Significant changes in leaf osmotic potential (increased by the parasitization), hormones (reduction in ABA production) and total soluble protein content (reduced in parasitized plants) as well as alterations in both fresh and dry biomass were observed.


In conclusion, we hypothesize that Cuscuta parasitization induces an increase in internal host plant defenses (primary metabolites fundamental for plant survival) at the expense of external ones (secondary metabolites) in the host species Artemisia, thus limiting plant defense against further parasitization.

"1H NMR Metabolic Profiling of Wild Ruminant Bighorn Sheep to Inform Wildlife Conservation Efforts." (53)


Galen O'Shea-Stone, Montana State University (Primary Presenter)

Valerie Copie, Montana State University


Wild ruminants are fundamental components of healthy ecosystems and of these, Bighorn sheep (O. canadensis) are iconic in Western North America. This species has experienced drastic declines in numbers due to overharvest, anthropogenic alterations of landscapes, as well as competition and comingling with domestic livestock over the last century. Limited tools are currently available to wildlife scientists to assess the nutritional health, disease status of bighorn sheep in the wild and the impact of grazing of domestic livestock on bighorn sheep habitats. The rather subjective and qualitative assessment tools currently used by wildlife biologists are thus inhibiting the understanding of wildlife-habitat relationships and the etiology of respiratory disease, which are major factors affecting the health and demographic vigor of bighorn sheep. Herein, 1H NMR metabolomics has been employed to better understand the impact of several practices that are used to monitor bighorn sheep health. Our overarching goal is to establish a new suite of tools for assessing the physiological status of this important wildlife species to advance our ecological understanding and enhance conservation.

As free ranging wild animals must be captured to obtain biological samples for analysis, this study aims to examine the metabolic changes that occur based on different capture techniques (helicopter, dart, or net) on wild bighorn sheep, as well as the nutritional state based on time of year capture took place. To achieve this goal, untargeted 1H NMR metabolomics was used to analyze the serum of 586 individual animals captured by different techniques, and under differential nutritional distress. The differential metabolite profiles of these different animal groups has enabled us to identify characteristic metabolites that separate these wild animals based on capture techniques and nutritional status. This study provides guidance for further studies, presenting evidence that difference capture techniques impact significantly the serum metabolomes of these animals. Furthermore, our data demonstrate the utility of employing metabolomics in wildlife ecology and conservation, providing a first example of its applications to the study of wild ungulates.

"Temporal changes in the metabolite profile of a Great Lakes invasive freshwater mussel related to environmental and biological factors" (103) - Zoom Session


Amanda Bayless, National Institute of Standards and Technology (Primary Presenter)

Ashley Elgin, NOAA Great Lakes Environmental Research Laboratory

Erik Davenport, NOAA National Centers for Coastal Ocean Science

Annie Jacob, Consolidated Safety Services-Dynamac, Virginia

Ed Johnson, NOAA National Centers for Coastal Ocean Science

Tracey Schock, National Institute of Standards and Technology


The Great Lakes Basin is the largest body of fresh water in the world, containing 21 % of the world’s supply of surface fresh water, and supports fishery and recreational industries worth $7 billion and $52 billion, respectively. This significant resource is monitored by the Great Lakes NOAA Mussel Watch Program, which employs freshwater mussels as bioindicators to assess contaminants and environmental change. Recently, NMR metabolomics was successfully applied to examine changes in the metabolome in the dreissenid mussel in response to pollutants, but it is not known whether the differences detected between sites are solely related to the presence of chemical pollutants or whether these detected differences are also related to environmental and biological factors that change throughout a season. This study aimed to identify temporal changes in the dreissenid mussel metabolic profile by measuring the metabolites in whole mussel tissue collected from a single site at the mouth of Muskegon Lake near Lake Michigan, USA over a 7-month period. Metabolites measured using NMR spectroscopy significantly fluctuated throughout the sampling period, but linear temporal trends were not observed. However, fluctuations in some metabolites were positively correlated with biological (mussel body index) and biotic (chlorophyll a, cyanobacteria pigment) measurements. All significant spectral features identified as glycerol were correlated to mussel body index, while alanine and succinic acid were correlated with chlorophyll a, and putrescine was correlated with cyanobacteria pigments. Additional measurements taken during this time period can be used to further associate these changes with either abiotic (temperature, DO), biological (reproductive status, DNA damage, cellular biomarkers), or chemical contaminant (pharmaceuticals, pesticides, alkylphenols) influences. This study helps us better understand how biological and environmental factors influence the mussel metabolome, which will improve the ability to detect and identify biological changes related to contaminant exposure and further support the use of metabolomics as a biomonitoring tool.

"Untargeted MSn-based Monitoring of Glucuronides in Fish: Screening Complex Mixtures for Contaminants with Biological Relevance" (115)


Jonathan D Mosley, Center for Environmental Measurement and Modeling, US Environmental Protection Agency (Primary Presenter)

Marina Evich, Center for Environmental Measurement and Modeling, US Environmental Protection Agency

Ioanna Ntai, Thermo Fisher Scientific

Jenna Cavallin, Center for Computational Toxicology and Exposure, US Environmental Protection Agency

Dan Villeneuve, Center for Computational Toxicology and Exposure, US Environmental Protection Agency

Gerald Ankley, Center for Computational Toxicology and Exposure, US Environmental Protection Agency

Tim Collette, Center for Environmental Measurement and Modeling, US Environmental Protection Agency

Drew Ekman, Center for Environmental Measurement and Modeling, US Environmental Protection Agency


The increasing complexity of contaminant mixtures in surface waters has made traditional assessments of risk to human health and the environment progressively more difficult, requiring new and advanced, novel analytical approaches. For example, methods are needed to identify contaminants for which routine chemical analyses are not performed, in addition to prioritizing both detected and undetected chemicals by both biological and toxicological relevance. Tracking biotransformation products of chemical contaminants in an untargeted fashion facilitates the identification of xenobiotics taken up by the resident environmental species (e.g., fish), and can inform risk based on known detoxification pathways. As a first step, we investigated xenobiotic glucuronidation, which is generally accepted as the most biologically relevant phase II metabolic pathway for pharmaceuticals, pesticides, and other environmental contaminants frequently encountered by organisms in the environment. Glucuronidation consists of the attachment of a glucuronic acid moiety to the xenobiotic (or to its phase I metabolite(s)) to aid excretion and thus, we developed a novel untargeted mass spectrometry-based method for identifying these glucuronides in tissues and biofluids in aquatic species. The application of this method to gallbladders from fish exposed to wastewater treatment plant effluent revealed the presence of over 70 biologically relevant xenobiotic-glucuronides, the majority of which were not targets of conventional contaminant monitoring. These results highlight the need for developing novel, biologically based untargeted screening methods for evaluating chemical contaminants in complex environmental mixtures to aid in both human and ecological risk assessment.

"A Systems biology Approach to Understanding the Stability of Peatland Carbon Pools under Climate Change" (127)


Roya AminiTabrizi, PhD Student (Primary Presenter)

Malak Tfaily, Assistant professor


Peatlands are organic-rich wetlands that act as globally important carbon sinks. Warming-induced environmental changes can accelerate the decomposition of carbon in these systems to become a major source of greenhouse gas. The drivers of peat decomposition, however, are poorly understood. In this study, peat from the S1-Bog at the Marcell Experimental Forest was aerobically incubated under three different temperatures (4, 21, and 35 oC) in a two-month long experiment. CO2 fluxes and d13C-CO2 were measured at five different time points. Untargeted metabolomics using direct inject mass spectrometry was conducted to study the effect of increasing temperature in the presence of O2 on overall metabolite composition in samples after each measurement time point. The activities of four extracellular enzymes, including β-glucosidase, Phosphatase, N-acetylglucosaminidase, and Leucine aminopeptidase were assayed to monitor carbon, phosphorus, and nitrogen cycles throughout the incubation period. Bacterial 16S rRNA gene amplicon paired-end sequencing was applied to assess the composition and spatiotemporal patterns of microbial communities. Shifts in metabolite diversities and composition, as well as abundances of metabolic pathways suggested that with increased temperature, in the presence of O2, microbes utilize various compounds where thermodynamic limitation does not necessarily play an important role. Activity of enzymes responsible for releasing N through various hydrolyses mechanisms increased towards the end of experimental period, suggesting that microbes were experiencing availability limitations. Distinct spatial patterns linked to increased temperature were observed within the microbial communities, with a prevalence of proteobacteria phylum and Rhizobiales order. Functional analysis of 16S rRNA metagenomes also revealed shifts in abundances of various metabolic pathways driven by both time and temperature. These results suggest that increased temperatures in peatlands destabilizes the carbon in recalcitrant peat through mineralization of organic matter and shifts microbial communities towards organisms that have developed physiological and metabolic adaptations to higher temperatures. This study will help us to understand potential changes in peatland distribution in response to future warming.

"Analysis of Geosmin Degraded by Bacterial Cultures by HSPME-GC-MS/MS" (131)


Trevor Andrew Johnson, University of Alberta (Primary Presenter)

Paulina de la Mata, University of Alberta

James Harynuk, University of Alberta

Camilla Nesbo, University of Alberta


A common culprit of the earthy-musty associated with rain and freshwater is geosmin, which is produced and emitted by Streptomyces Spirulina in water. This compound, though non-toxic, has been of concern for perception of quality in food, drink, and other sensory-based industries. Geosmin is commonly associated with the smell of rain and has a very low threshold of detection by the human nose at <10 ng/L. There are few methods for the removal of geosmin from water, and the process often involves extensive filtration and pretreatment. One potential way to remove geosmin from water may be by the growth of geosmin-metabolizing bacteria. By growing select bacteria in a culture with a supply of geosmin and analyzing the amount of the compounds remaining, we can gain insight into the efficacy and potential of various bacteria for removal of geosmin. In this study, water samples containing high levels of geosmin were exposed to bacteria other than S. Spirulina, and cultured. Samples from each culture were collected at multiple time points after the addition of geosmin and analyzed by headspace SPME followed by GC-MS/MS analysis, and compared to control samples (geosmin, water, and no bacteria). The method used can detect geosmin in water at levels below the human nose detection threshold. Preliminary results have shown that compared to control samples, bacterial growth (high or low) had minimal impact on levels of geosmin over time.

"Metabolomics based analysis for identifying metabolites to differentiate fungal phenotypes" (159)


Chathuri Udeshika Gamlath Mohottige, Mississippi state university (Primary Presenter)


Metabolomics is a powerful tool for investigating gene function because it can provide information on the presence and concentration of metabolites thereby differentiating phenotypes at the molecular level. In this study, the untargeted metabolomics workflow has been used to study the fungus Macrophomina phaseolina which belongs to division ascomycetes. This fungus is one of the most disastrous pathogens to a broad range of crops with a worldwide distribution. There are two main phenotypes of this fungus, namely fluffy and flat. These strains can be easily identified even with the naked eye by grown on the PDA plates. The pathogenicity of the two phenotypes is different. The objective of this study is to identify different types of metabolites or levels of metabolites to differentiate two phenotypes. In order to identify the mechanisms involved in the phenotype differentiation, volatiles evolved from the cultures of two morphologies were analyzed by using a headspace SPME based GC-MS technique and followed by chemometric analysis. The analysis of the volatile and semivolatile metabolites could lead to understanding the mechanisms involving the pathogenicity of this fungus. Ultimately, a sensory device can be developed to diagnose early fungal infection. The key metabolites identified from the untargeted metabolomics approach and their important pathways will be presented.

COVID 19 and Viral Infection Metabolomics

"The effect of Zika virus infection on urine metabolome of pregnant women: a longitudinal study" (169)


Sicong Zhang, The University of Georgia (Primary Presenter)

Dorothy Ellis, Department of Biostatistics, College of Public Health & Health Professions and College of Medicine, University of Florida, Gainesville, FL, 32611

Jacquelyn Walejko, Department of Biochemistry & Molecular Biology, University of Florida, Gainesville, FL, 32610

Maria Arévalo, Center for Vaccines and Immunology, University of Georgia, Athens, Georgia, USA 30602

Ted Ross, Center for Vaccines and Immunology, University of Georgia, Athens, Georgia, USA 30602

Susmita Datta, Department of Biostatistics, College of Public Health & Health Professions and College of Medicine, University of Florida, Gainesville, FL 32611

Arthur S Edison, University of Georgia


Zika virus (ZIKV) infection in pregnant women can lead to stillbirth and increased incidence of fetal encephalopathy. Despite the popularity of studies in this area recently, there is currently still no sufficient treatment for it. We here conducted an untargeted metabolomics study on ZIKV infected versus non-infected mothers to aid in etiological mechanism understanding. We recruited 275 pregnant women from Puerto Rico in an IRB approved study. Metabolomic analysis was conducted on urine specimens from 10 ZIKV-infected women and 10 control women matched by gestational age and maternal age. Samples were collected monthly after patients’ first prenatal care visits at up to 6 time points and then analyzed by proton Nuclear Magnetic Resonance (NMR) spectroscopy. We manually bucket 649 features from processed one-dimensional NMR spectra and analyzed them in nonparametric mixed effect models independently to compare the overall behavior of each NMR feature overtime between Zika-positive and control patients. Fifty features were found statistically significant (FDR-corrected p-value < 0.15) between the two groups and 38 of them were identified. Three metabolites including trigonelline, 3-aminoisobutyrate and 3-methyhistidine have higher concentrations in Zika-positive samples while 12 metabolites are more abundant in control samples. These metabolomic changes may lead us to a better understanding of mechanisms that cause poor fetal outcomes as well as effects of virus infection on human pregnancy.


"Combination of Ozone-Induced dissociation with DESI imaging mass spectrometry to visualize double bond positional isomers of phospholipids." (85)


Mark Towers, Waters Corporation

Lisa Reid, Waters Corporation

Martin Green, Waters Corporation

Emmanuelle Claude, Waters Corporation (Primary Presenter)


Ozone-induced dissociation (OzID) was first implemented analyzing lipid extracts by ESI-MS to assign sites of unsaturation in complex lipids1. Understanding the profiles of metabolic changes within the tissue can be challenging due to the presence of numerous isobaric and near isobaric none mass resolved species. DESI imaging is fast becoming technique used to map small molecules such as lipids directly from biological samples, without the need to extensive sample preparation. Here we demonstrate the implementation of DESI in combination with OzID to determine the double bond position of isobaric lipids within an imaging experiment.

Data were acquired using a DESI source (Prosolia) which was mounted on a SYNAPT HDMS G2-Si mass spectrometer (Waters). DESI spray conditions were set at 2µl/min, 98:2 MeOH: water with nebulizing gas pressure of 1-2 bar. Images were visualized in HDI 1.5 (Waters) and image correlation performed.

The travelling wave ion mobility separation enabled Q-ToF instrument was modified to allow ozone to be introduced into the ion mobility cell instead of nitrogen. The time ions spent within the mobility cell could be controlled by reducing the wave height of the traveling wave. Furthermore, the SYNAPT G2-Si has collision cell prior to the place where the OzID occurs, allowing CID fragmentation prior to OzID.

DESI imaging experiments have been performed mouse brain sections. Various lipid peaks have been isolated and subjected to OzID in profiling and imaging mode. Spectra of varying degrees of complexity have been generated, from isolated lipids with a single double bond to multiple isobaric lipids with multiple double bonds. As an example, lipid at m/z 810.6 was isolated and subjected to OzID during a DESI imaging experiment. Several neutral loss peaks were detected: a pair of 66&82 Da corresponding to an n-7 monounsaturated acyl chain and a pair of 94&110 Da corresponding to an n-9 monounsaturated acyl chain. In addition, the same tissue section was re imaged with CID prior to OzID. It was possible to determine that at least two sodiated lipids were present with a combination of 16:0_20:1 and 18:0_18:1 fatty acid chains. Finally fragments indicative of sn-1on the glycerol backbone positions were detected, such as m/z 379.3 for sodiated fatty acid chain16:0, and m/z 407.3 for fatty acid chain 18:0.

Incorporation OzID into a DESI imaging experiment of m/z 810.6, it was possible to visualize at least two different monounsaturated acyl chain n-7 and n-9 with significantly different distribution. Furthermore, with the CID fragmentation prior to OzID, experiment, it was possible to fully determine the structure of the sodiated isomer lipids being PC 16:0/20:1(n-7) and PC 18:0/18:1(n-9), based fragment ion distributions.


1 Thomas et al. Anal. Chem. 2008, 80, 303.


"Pediatric Multi-Organ Dysfunction Syndrome: Untargeted Lipidomic Analysis" (13)


Mara L. Leimanis Laurens, PhD, Pediatric Critical Care Unit, Helen DeVos Children's Hospital, 100 Michigan Street, NE, Grand Rapids, MI, 49503 (Primary Presenter)

Karen Ferguson , Helen DeVos Children's Hospital

Emily Wolfrum, Van Andel Institute, Bioinformatics & Biostatistics Core, Grand Rapids, MI, United States.

Brian Boville, Pediatric Critical Care Unit, Helen DeVos Childrens Hospital, Grand Rapids, MI; Department of Pediatric and Human Development, Michigan State University, East Lansing, MI

Dominic Sanfilippo, Pediatric Critical Care Unit, Helen DeVos Children’s Hospital, Grand Rapids, MI, United States.

Jeremy Prokop, Department of Pediatric and Human Development, College of Human Medicine, Michigan State University, Life Sciences Bldg. 1355 Bogue Street, East Lansing MI 48824

Todd Lydic, Department of Physiology, Collaborative Mass Spectrometry Core, East Lansing, MI, United States.

Surender Rajasekaran, Pediatric Critical Care Unit, Helen DeVos Children's Hospital, 100 Michigan Street, NE, Grand Rapids, MI, 49503


Background: Lipids are stable molecules involved in signaling, metabolism and inflammation, thereby representing attractive biomarkers for use in the intensive care environment. We investigated if lipid classes within the plasma lipidome corresponds to severity and nutritional status in the critically-ill child.

Methods: Children with multi-organ dysfunction syndrome (MODS) (n=28) and sedation controls had blood draws at three time points. In the cohort with MODS (n=24), there were (8) patients that needed veno-arterial extracorporeal membrane oxygenation (VA ECMO), patients with a more severe form of MODS that requires mechanical assistance to survive. Following separation, blood plasma lipid profiles were determined by nano-electrospray (nESI) direct infusion high resolution/accurate mass spectrometry (MS) and tandem mass spectrometry (MS/MS) and compared to nutritional profiles and PEdiatric Logistic Organ Dysfunction (PELOD) extracted from medical records.

Results: PELOD scores were not significantly different between MODS and ECMO groups across all time points (Global F-stat, p = 0.66), which were logically higher than sedation controls. Lipid profile provided separation between sedation controls and MODS patients including lysophosphatidylserine (lysoPS) (P – value = 0.004), total phosphatidylserine (PS) (P – value = 0.015), and total ether-linked phosphatidylethanolamine (PE) (P-value = 0.03), from a generalized linear model. Phospholipids and NEFA levels in patients needing ECMO were closer in value to sedation controls and significantly different from MODS patients without need for ECMO (Welch-test for unequal variances, P < 0.0001; P = 0.0030, respectively). Dietary intake analysis revealed changes in lipid profiles that corresponded to percent calories and protein intake.

Conclusion: Lipids measurement in the intensive care environment shows dynamic changes in patients with multiple organ dysfunction relative to controls except for those patients that require ECMO therapy, suggesting a novel area of future exploration of biomarkers for defining critically ill children.

"Plasma lipidomics profile is linked to the presence and number of obstructed epicardial coronary arteries in adults" (113)


Manuel Garcia-Jaramillo, Oregon State University (Primary Presenter)

Armando Alcazar Magana, Oregon State University

Elizabeth Le, OHSU

Gerd Bobe, Oregon State University

Donald B. Jump, School Biological and Population Health Sciences, Oregon State University

Claudia S. Maier, Oregon State University

Nabil Alkayed, OHSU

Sanjiv Kaul, OHSU


Background: Coronary artery diseases (CAD), characterized by lipid plaque formation in the coronary artery intima, is the most common cardiovascular disease (CVD), accounting for approximately 30% of deaths worldwide. Lipidomics is an emerging methodology that allows for comprehensive identification of individual lipids and their fatty acid composition in complex biological samples. Moreover, lipidomics can provide information about the etiology of CAD and identify potential CAD biomarkers for patient stratification and for new pathway-specific therapies.

Objective: In this report, we tested whether presence and number of obstructed epicardial coronary arteries were linked to specific lipid classes and species in adults.

Methods: Using ultra-high-performance liquid chromatography (UHPLC) coupled to a quadrupole time-of-flight (TOF) mass spectrometer, we conducted untargeted lipidomic analysis in plasma samples obtained from 112 individuals, (61 males/51 females), ranging in age from 18 to 87. Relative concentrations of 460 annotated lipids were linked to prevalence and number of obstructed (70-100% stenosis) epicardial coronary arteries, as determined by coronary angiography, in the subjects (number of obstructed arteries: none, n=38; 1, n=31; 2, n=23; 3, n=20). Raw data from untargeted analysis was imported into MS-DIAL (v. 2.80) software for lipidomics analysis. Data were statistically evaluated using SAS version 9.2.

Results: Overall, the presence of obstructed epicardial coronary arteries was not significantly associated with changes in total lipid abundance. However, multiple lipid classes were altered dependent on CAD severity. Of the lipid groups, non-esterified fatty acids (FFA; -86%; P = 0.03), acetyl-carnitines (AC; -28%; P = 0.009), cholesterol esters (CE; -16%; P = 0.01), sphingomyelins (SM; -14%; P = 0.02), lysophospholipids (Lyso-PC; -23%; P = 0.05), phosphatidylethanolamines (PE)-plasmalogen (PEP; -17%; P = 0.002), PE (-17%; P = 0.02), glycerophosphates (PA; -14%; P = 0.03), and phosphatidylcholines (PC; -8%; P = 0.02), were significantly decreased in patients with 3 vs. 1 obstructed epicardial coronary artery. Whereas, FFA and most mono-, di-, and triacylglycerol species were elevated in participants with vs. without CAD, phospholipids, especially those with phospho-ester and ether bonds, exhibited lower levels in individuals with CAD. Interestingly, precursors of pro-inflammatory oxylipins, e.g., arachidonic acid (ARA; C20:4, ω6) were higher in patients with CAD, and precursors of anti-inflammatory oxylipins, e.g., docosahexaenoic acid (DHA; 22:6, ω3), were inversely associated with the severity of CAD. In addition, DHA-containing CE were also inversely associated with CAD severity. Non-esterified fatty acids (7 of 12 annotated FFA) were elevated in patients with one obstructed epicardial coronary artery but, surprisingly, FFA levels decreased by at least 35% with increasing numbers of obstructed arteries.

Conclusions: The outcome of these studies suggest that the analysis of plasma lipids may assist in identifying individuals who will warrant further cardiac evaluation of CAD burden.

Monday, September 14, 6:05-6:50 EDT

Biomedical applications, from disease to exposome

"Machine Learning Predicts Renal Cell Carcinoma Status from Urine Using Multiplatform Metabolomics" (32)


Olatomiwa Bifarin, The University of Georgia (Primary Presenter)

David Gaul, School of Chemistry and Biochemistry, Georgia Institute of Technology

Samyukta Sah, Georgia Institute of Technology

Rebecca Arnold, Department of Urology, Emory University, Atlanta, GA 30342

John Petros, Emory University

Facundo M. Fernández, School of Chemistry and Biochemistry, Georgia Institute of Technology

Arthur S Edison, University of Georgia


Due to the characteristic asymptomatic progression of Renal Cell Carcinoma (RCC), an early diagnosis of RCC greatly improves survival. Currently, RCC is detected through cross-sectional imaging, followed by renal mass biopsy, which is invasive and riddled with sampling errors. Hence, there is a critical need for a non-invasive diagnostic assay. RCC is a disease of altered cellular metabolism with the tumor(s) in close proximity to the urine in the kidney suggesting metabolomic profiling would an excellent choice for assay development. We applied machine learning (ML) techniques to predict RCC status from an integrated liquid chromatography/mass spectrometry (LC/MS) and nuclear magnetic resonance (NMR) metabolic profile. NMR spectroscopy and LC/MS experiments were collected for urine from 82 RCC patients and 174 healthy controls. The cohort was divided into two sub-cohorts for training and validation purposes. Discriminatory 1H NMR and MS features were selected using a hybrid-based feature selection pipeline applied to the training cohort. Three ML techniques with different induction biases were used for training and hyperparameter tuning. Final assessment of RCC status prediction was made using the selected features and the tuned ML algorithms on the test cohort. The model cohort consists of a pair of 31 healthy controls and RCC samples selected using a propensity score matching algorithm (PSM). After PSM, the cohort were gender matched (17 males, 14 females), and had statistically insignificant difference in age (Student t-test p-value=0.64) and BMI (Student t-test p-value=0.06). Smoking history and race’s statistics also improved considerably when compared to the pre-matched cohort. In addition, all RCC stages were represented – early stage RCC (I/II) represents 55% while the late stage RCC (III/IV) represents 45% of the model cohort. Using this model cohort, 49 NMR features and >6000 LC-MS features were filtered to 132 features using the following methods: >2-fold difference between groups, q <0.05 (Student t-test, FDR Benjamin-Hochberg), and removal of highly correlated features (Pearson's correlation coefficient, r >0.8). Next, we used a random forest recursive - feature elimination (RF-RFE) technique and partial least square regression discriminant analysis (PLS-DA) to select the top 20 discriminatory features, using Gini index and VIP scores respectively. The ten metabolic features that overlapped between the two sets was selected as the metabolic panel. All ten features are from LC/MS. The panel were used to train and tune selected hyperparameters in Random Forest (RF), K-nearest neighbor (k-NN), and support vector machine (SVM-Lin and SVM-RBF) classifiers in the model cohort. In the test cohort (healthy controls=143, RCC=51), the best ML model using Random Forest classifiers, discriminated RCC and healthy controls with an accuracy, sensitivity, and specificity of 98% under cross-validation conditions. 2-aminoacetophenone was one of the tentative metabolic biomarkers discovered among the selected features, as the identification of the MS metabolic panel is underway. Similar analysis on NMR features led to the selection of a 4-metabolite panel: hippurate, 1-methylnicotinamide, mannitol, and lactate. However, with a lower classification accuracy of 87% in the test cohort. Our results provide evidence that RCC diagnosis may be possible via a routine urine test.

"Fluoroacetate Toxicity and Metabolism in the Adult Zebrafish (Danio rerio)" (44)


Casey A Chamberlain, Washington University in St. Louis (Primary Presenter)

Madelyn Jackstadt, Washington University in St. Louis

Gary Patti, Washington University in St. Louis


Fluoroacetate (FA), also known by its brand name “1080”, is a highly lethal compound with no known antidote and applications ranging from public health to environmental protection. Created in the 1940s under the U.S. Department of the Interior during an initiative to generate new pesticides, FA was once widely used in agriculture throughout the world. However, since the 1970s, its use and production have been largely restricted in the United States due to its environmental impact and danger to human and animal health. FA is toxic to obligate aerobic organisms at low doses (humans 2-10 mg/kg) due to its metabolism into fluoroacetyl-CoA and, subsequently, fluorocitrate (FC), a “suicide substrate” which strongly binds and competitively inhibits aconitase. FC binding disrupts the tricarboxylic acid (TCA) cycle by preventing the conversion of citrate to isocitrate, causing excess citrate accumulation, depletion of downstream TCA intermediates, inhibition of the flow of electrons needed for oxidative phosphorylation, and ultimately leading to reduced cellular ATP production. In humans, acute FA poisoning causes cardiac (ventricular fibrillation, hypotension) and neurological (seizures, coma) complications within hours of lethal exposure, and treatment is usually limited to symptom management and supportive care.

Despite decades of dedicated research using a variety of cell culture and animal models, there is currently no biomarker or effective antidote for FA toxicity. Furthermore, due to its complete lack of sensory recognition (tasteless, odorless, colorless solution), low production cost, and high lethality to humans, there is significant interest in FA from a homeland security, public health, and public safety perspective regarding its potential use as a chemical warfare or terrorism agent. Hence, there is a pressing need for novel approaches to examine and combat FA toxicity. We addressed this need by performing the first investigation of FA toxicity and metabolism in adult zebrafish (ZF) (Danio rerio). ZF hold key advantages over other experimental animal models that allow for a more streamlined approach to both toxicology and metabolism-based drug screenings. We conducted a UPLC-HRMS-based systemic multi-organ metabolomic analysis of ZF conditioned in various concentrations of FA, assessing what we term the “FA response” including detection of fluoroacetate and fluorocitrate, elevated citrate, reduced TCA cycle intermediates subsequent to citrate, and depleted ATP, among other effects, in treated ZF compared to controls. Additionally, we tested and observed metabolomic alterations using other FA-like drugs that are commonly used in FA toxicology, including methyl-FA, fluoroethanol, and fluoroacetamide, and observed differential tolerance associated with these drugs both systemically and in individual organs. Lastly, we describe preliminary investigations of the efficacy of select antidote candidates previously tested in other animal models in the literature and our evaluation of their ability to reverse the FA response.

"Exposure to inorganic arsenic and its methylated metabolites alters metabolic profiles in INS‑1 832/13 insulinoma cells and isolated pancreatic islets" (62)


Yuanyuan Li, University of North Carolina at Chapel Hill (Primary Presenter)

Christelle Douillet, University of North Carolina at Chapel Hill

Madelyn Huang, National Toxicology Program, NIEHS

Rowan Beck, University of North Carolina at Chapel Hill

Susan Sumner , UNC Chapel Hill

Miroslav Styblo, University of North Carolina at Chapel Hill


Inorganic arsenic (iAs) is a diabetogen, but mechanisms underlying its diabetogenic effects are poorly understood. Exposure to arsenite (iAsIII) and its metabolites, methylarsonite (MAsIII) and dimethylarsinite (DMAsIII), have been shown to inhibit glucose stimulated insulin secretion (GSIS) in β-cells and isolated pancreatic islets. GSIS is regulated by complex mechanisms that depend on metabolism of glucose and other energy producing pathways. The present study used untargeted metabolomics approach to identify metabolic profiles and pathways that were perturbed in cultured INS-1 832/13 rat insulinoma cells (β-cells) and isolated murine pancreatic islets after exposures to iAsIII, MAsII, and DMAsIII. We found that the GSIS in islets was significantly inhibited by all three arsenicals; while the GSIS in β-cells was inhibited by DMAsIII with statistical significance. Supervised orthogonal partial least squares discriminate analysis revealed that exposure of DMAsIII treatment resulted in major metabolic perturbations in β-cells affecting 37 metabolites, as compared to 9 and 5 metabolites perturbed by iAsIII and MAsIII, respectively. Two metabolites, acetylcarnitine and succinic acid, were decreased following exposure to each of the three arsenicals. Pathway analysis revealed 25 metabolic pathways enriched by treatment with DMAsIII, including multiple pathways of amino acid transport and metabolism. Six and 17 pathways were enriched by exposure to iAsIII and MAsIII, respectively. The D-glucuronic acid pathway was the only pathway downregulated by all three arsenicals. In spite of their significant effects on GSIS, overall impact of arsenical exposures on metabolome of islets was less than that in β-cells. Several metabolites were altered by exposure to one or more arsenicals in both β-cells and islets, including acetylcarnitine, glutamate, suberic acid, glutathione and ornithine. The pathways of carbohydrate and amino acid metabolism were the pathways most affected in both β-cells and the islets. While each of the three trivalent arsenicals perturbed specific metabolic pathways, which may or may not be associated with GSIS, all three arsenicals appeared to impair mechanisms that support ATP production or antioxidant defense in mitochondria. These results suggest that impaired ATP production and/or mitochondrial dysfunction caused by oxidative stress may be the mechanisms underlying the inhibition of GSIS in β-cells exposed to trivalent arsenicals.

Results of this study will inform future mechanistic research to better understand how iAsIII exposure impairs GSIS in insulin secretion cells from pancreatic islets; which might provide guidance for discovery of nutritional treatment to reduce the risk of arsenic exposure.

"3D spatial metabolomics (“chemical cartography”) identifies determinants of infectious disease tropism and infection tolerance" (68)


Laura-Isobel McCall, University of Oklahoma (Primary Presenter)

Ekram Hossain, University of Oklahoma

Sharmily Khanam, University of Oklahoma

Danya Dean, University of Oklahoma

Gautham Gautham, University of Oklahoma

Chaoyi Wu, University of Oklahoma

Sharon Johnson, University of California, San Diego

Diane Thomas, University of California, San Diego

Shelley Kane, University of Oklahoma

Adwaita Parab, University of Oklahoma

Karina Flores, University of Oklahoma

Mitchelle Katemauswa, University of Oklahoma

Camil Gosmanov, University of Oklahoma

Stephanie Hayes, University of Oklahoma

Yiming Zhang, University of Oklahoma

Jair Siqueira-Neto, University of California, San Diego

Danyang Li, Beijing Normal University

Christine Woelfel-Monsivais, University of Oklahoma

James McKerrow, University of California, San Diego

Pieter Dorrestein, University of California, San Diego

Krithi Sankaranarayanan, University of Oklahoma


Two central questions of infectious disease pathogenesis are: 1) Why the outcome of infection in different individuals varies, even though they are exposed to the same pathogen load; and 2) Why a given disease happens in specific tissue locatons (disease tropism). To address these questions, we leverage a new integration of untargeted liquid chromatography-tandem mass spectrometry (LC-MS/MS)-based metabolomics, 3D spatial modeling and infection biology (“chemical cartography”). As an example of the power of this approach, we used chemical cartography to determine sites of disease tropism during Trypanosoma cruzi infection. T. cruzi is an unicellular parasite that is the cause of Chagas disease, infecting over 7 million people worldwide. Chagas disease symptoms include cardiac arrhythmias, cardiac apical aneurysms, enlargement of the oesophagus and of the colon (megaoesophagus and megacolon). By systematically mapping infection-induced metabolic perturbations in the heart and gastrointestinal tract, we demonstrated that infection is associated with persistent alterations in overall metabolism at sites of Chagas disease tropism: heart apex, oesophagus and large intestine, even after parasite load is decreased by the immune system. Strikingly, these metabolically-perturbed sites are also distinct from the tissue sites of highest parasite load. In contrast, metabolism at sites that are not associated with disease tropism, such as the small intestine, goes back to normal once parasite levels decrease. Metabolic families altered by infection include acylcarnitines and phosphocholines. Experimental treatment with carnitine abrogated infection-induced acute stage mortality, by selectively restoring overall cardiac and plasma metabolism. Carnitine-treated, T. cruzi-infected animals present the same parasite load as vehicle-treated controls, but a cardiac and plasma metabolic profile overall comparable to uninfected animals, via “re-setting” of fatty acid and amino acid metabolism. This metabolic restoration was accompanied by reduction in cardiac strain (cardiac Bnp gene expression). Overall, our results demonstrate a new, metabolomics-based method to identify sites of infectious disease tropism. Furthermore, our metabolomics data enabled us to discover a new approach to treat Chagas disease and generated a new, metabolism-based perspective on Chagas disease tolerance. Current work is demonstrating the broad applicability of this approach across infectious diseases.

"Identification of MSC critical quality attributes through non-destructive, in-process characterization of cellular outputs" (70)


Xunan Shen, UGA (Primary Presenter)


Mesenchymal stem cells (MSCs) have potential in stem cell-based therapies for tissue repair, organ transplantation and the treatment of autoimmune disease. However, unintended differentiation and unpredictable performance of transplanted MSCs in human present challenges in the clinical application of MSC-based therapies. We are using metabolomics, other phenotypic measurements, and modeling to discover critical quality attributes (CQAs), which could be used to predict the efficacy and potency of MSCs for transplantation. In this poster, we will show our results using NMR, MS and cytokines analysis to attempt to discover CQAs of human umbilical MSCs. MSCs endometabolites were characterized by 1H proton NMR and LC-MS/MS experiments after culturing. We will be correlating changes in the cellular metabolites with various phenotypic cell measurements. The goal is to determine CQAs that can predict the function of these MSCs.

"Measuring relevant markers in the place they matter: metabolomics of cerebrospinal fluid for better pediatric brain tumor therapy" (72)


Boryana Petrova, Boston Children's Hospital (Primary Presenter)

Naama Kanarek, Boston Childrens Hospital


Pediatric blood tumor patients with masses in their brain face grave prognosis and limited treatment options. One of the few effective chemotherapies, methotrexate is associated with toxicity and long-term neurocognitive side effects; 60% of pediatric leukemia survivors show symptoms of cognitive deficit following treatment. While methotrexate treatment results in systemic depletion of the essential vitamin folate, the relevant pool of folate important for brain-residing tumors is in the cerebrospinal fluid (CSF), not in blood or plasma. Unfortunately, our knowledge of the immediate metabolic effects of malignant brain lesions therapy in the CSF as well as our methods to study the broader metabolic impacts in this clinically relevant setting is rudimentary.

Using orthotopic pediatric-leukemia cell-derived xenografts mouse models we study cancer that localizes to the brain. We are optimizing an advanced mass spectrometry (MS) platform to monitor tumor progression and accompanying metabolic changes as well as consequences of methotrexate treatment in the relevant microenvironment – the CSF. We focus on several highly relevant metabolites including folate and folate metabolism intermediates, as well as oxidative stress markers. We perform measurements in the CSF and aim to establish an LC-MS-based platform to robustly detect clinically relevant metabolic changes with further focus on labile metabolites. Our LC-MS platform will allow the analysis of CSF samples from patients and mouse models to define metabolic changes that correlate with response to therapy, or harmful side effects.

To establish the method of metabolite profiling of freshly collected CSF in our lab, we performed several pilot studies that were designed to tackle some of the technical challenges in this study. In one pilot study we were able to detect methotrexate and a drop in folate in methotrexate-treated mice, but in parallel non-treated controls. In another pilot study we optimized a method that allowed us to measure oxidative stress induced by methotrexate in the brain that is thought to contribute to the infamous methotrexate brain toxicity common in pediatric leukemia patients.

This project will decipher the unknown metabolic changes in the brains of tumor-bearing mice following various therapy regimes, assess the extent of tumor inhibition by these regimes, and define the metabolic shifts reflecting the success of the treatment in mice and validated in patients. This study has the potential to significantly improve the treatment efficacy of methotrexate in young leukemia patients with the clinical goal of reducing toxicity in pediatric patients during chemotherapy, and ultimately preventing the development of chemotherapy-attributed long-term side effects

"The Metabolomic and Lipidomic Profile of X-Chromosome Deletion Disorder using UHPLC-HRMS" (98)


Hoda Safari Yazd, University of Florida (Primary Presenter)

Vanessa Y. Rubio, University of Florida

Casey A Chamberlain, Washington University in St. Louis

Richard A Yost, University of Florida

Timothy Garrett, Univ of FL-Pathology


Intellectual disorders connected to the X chromosome's deletion present a challenging task in discovering a connection between symptoms and the metabolome and phenotypic expression. One specific disease of X-chromosomal deletion, Fragile X syndrome, is the most frequent intellectual disabilities related to X chromosomal deletion. Several Metabolomic studies have been performed on different X-chromosomal deletion disorders using the maternal immune activation (MIA) model, Fragile X (FMR1 knockout) model, and a variety of other models to understand the dysregulation of metabolic pathways in rare X-chromosomal deletion disorders. Previous studies on X-linked intellectual disabilities have demonstrated promising metabolomic results based on mouse models; however, little metabolomic research has been performed using human-derived samples. These human studies are essential in helping to understand the metabolic signatures of X-linked intellectual disabilities and genetic deletion disorders in humans.

In this work, Neural deletion progenitor cells (NPC) were analyzed against normal neural progenitor cell samples as the control set (ten million cells per sample) and cell count used for pre-normalization. Extraction techniques were optimized for these specific NP cell lines to ensure full coverage of lipid lipids and metabolites and a Folch-based extraction method. The acquisition was conducted on a Thermo Q-Exactive mass spectrometer operating at 70 – 1,000 m/z in both positive and negative ionization. Chromatographic separation was performed using standard reverse-phase for metabolites and lipids. LipidMatch software employed to putatively identify lipids based on m/z and fragmentation spectra. MZmine was used for feature detection, integration, and peak list alignment for metabolite datasets. MetaboAnalyst 4.0 was employed for multivariate statistical analysis.

Lipidomics analysis found that there are few highly expressed lipid species containing very-long-chain fatty acids (VLCFA) with different degrees of unsaturation in the deleted X sample (like plasmenyl-PC(P-18:1/22:4), plasmenyl-PC(P-18:1/22:4), plasmanyl-PC(O-16:1/22:4), and plasmenyl-PC(P-16:0/22:4)). These fatty acids are usually correlated with peroxisomal β-oxidation defects and have been observed in many neurodegenerative disorders. Also, the PC/PE ratio is extremely altered in deleted X compared to control (PC/PE = 8.43 versus 6.32), this higher PC/PE ratio could be an indicator of lipid disequilibrium in deleted X samples. On the other hand, lipid enrichment analysis shows downregulation of PE biosynthesis, negative intrinsic curvature, endoplasmic reticulum, and mitochondrion lipid metabolism and upregulation of PG metabolism and long-chain polyunsaturated fatty acids in deleted X samples.

Metabolomics analysis results have revealed that NPs from deleted X samples had dramatic impacts on neurotransmitters, precursors, or catabolites. The most significant metabolites in the deleted X group that are related to neurotransmitters include glutamine, glycine, phosphocholine, 4-aminobutanoate (GABA), taurine, and aspartate. In addition, many of the other significantly identified metabolites were found to be nucleosides, nucleotides, or their derivatives, such as isocytosine, guanine, cytidine, adenine, and guanosine. Metabolite set enrichment analysis (MSEA) that performed on differential metabolites (p-value < 0.05%) screened in deleted X and control samples reveals that several amino acids and their metabolites were significantly changed in the deleted X including glutathione metabolism, alanine, aspartate, and glutamate metabolism, glycine, serine, and threonine metabolism, methionine metabolism, arginine and proline metabolism, glutamate metabolism and especially homocysteine degradation.

"Dysregulated nucleotide sugar metabolism in breast cancer" (126)


Shao Thing Teoh, Michigan State University (Primary Presenter)

Martin Ogrodzinkski, Michigan State University

Christina Ross, National Cancer Institute

Kent Hunter, National Cancer Institute

Sophia Lunt, Michigan State University


Changes in the glycosylation patterns of glycoproteins and glycolipids are associated with tumorigenesis, tumor development and metastasis. In all known cases, the sugar molecules involved in glycosylation are first activated by conjugation with specific nucleotides. Hence, nucleotide sugar biosynthesis is closely linked to the glycosylation ability of the cell. However, changes in nucleotide sugar metabolism have rarely been studied in the context of tumor development. Here, we present the results of metabolomic analyses implicating nucleotide sugar metabolism in murine breast cancer development and metastasis. Using mass spectrometry-based metabolomics, we initially investigated metabolic differences between PyMT-driven tumors with differing metastatic propensities. Our results pointed to an increase of intracellular N-acetylneuraminic acid (sialic acid) in highly-metastatic tumors compared to low-metastatic tumors. Using steady isotope labeling, we additionally found that highly-metastatic 4T1 mouse breast cancer cells have higher sialic acid metabolic flux and contain increased levels of CMP-sialic acid (the activated form of sialic acid) relative to syngeneic cell lines that are poorly- or non-metastatic. Further, knockout of Cmas, a key enzyme in CMP-sialic acid production, significantly attenuated in vivo metastatic ability in two highly-metastatic breast cancer cell lines of differing mouse genetic backgrounds and initiating oncogenic events, demonstrating the functional importance of sialic acid metabolism in breast cancer metastasis. Based on these results, we expanded our analytical coverage to include other nucleotide sugars. Comparison of PyMT-positive tumors versus normal mammary tissue reveal additional nucleotide sugars besides CMP-sialic acid are significantly elevated in tumor tissue. Ongoing work aims to 1) elucidate the functional importance of nucleotide sugar pathways perturbed in breast cancer using CRISPR/Cas9-mediated gene knockout, and 2) elucidate the functional role of CMP-sialic acid and other nucleotide sugars in breast cancer metastasis.

"Suitability of Dust and Domestic Sludge Standard Reference Materials (SRMs) for Metabolomics Studies" (128)


Kehau A. Hagiwara, National Institute of Standards and Technology (Primary Presenter)

Tracey Schock, National Institute of Standards and Technology

Jessica Reiner, National Institute of Standards and Technology


Interest in the Built Environment (BE) has accelerated in the recent COVID-19 climate. Prior to stay-at-home recommendations, Americans would spend an average of 90% of their time in enclosed spaces (home, offices, cars, etc.). While this fueled studies into the relationships between the BE and human health in the last 20 years, the COVID-19 situation has emphasized the need to understand the interplay between our health and environment. Specifically, there is an emerging area of comprehensive chemical analyses, including metabolomics, focused on better understanding chemical transformations and or markers that occur within the BE and how they impact materials and humans. These measurements have the potential to affect forensic analyses, community health efforts, engineering controls, etc. and further elucidate the relationships between the BE and human health. Recent studies have highlighted the opportunities and knowledge gaps for new measurement techniques to better understand the intricate interactions between humans and their associated environments. One of the most important areas to address is the need for quality control (QC) mechanisms and materials to facilitate harmonization across the field.

The National Institute of Standards and Technology (NIST) has a variety of existing Standard Reference Materials (SRMs) that may suit the needs of researchers interested in BE-related metabolomics analyses, specifically house dust (SRM 2585) and domestic sludge (SRM 2781). These materials are complex, heterogeneous samples that are relevant to targeted BE sample types and contemporary metabolomics questions. These materials, if deemed fit for metabolomics purposes, would provide a market-ready solution to important workflow development processes and QC needs.

Consequently, this study investigates the suitability of SRMs 2585 and 2781 for metabolomics use. Proton NMR spectroscopy, known for its robustness and high reproducibility, was used to analyze polar constituents extracted from the materials. Metabolite profiles were generated for each of the SRMs and compound identification was completed when possible. These samples were also assessed for technical variability. This work offers preliminary support for the incorporation of existing SRMs into metabolomics studies to address measurement QC and study comparability leading towards harmonization in BE research.

"Exploring the Gut-Brain-Liver Axis in Familial Dysautonomia: How Microbes and Metabolism Impact Neurodegeneration" (136)


Alexandra Marie Cheney, Montana State University (Primary Presenter)

Stephanann Costello, Montana State University

Valerie Copie, Montana State University

Frances Lefcort, Montana State University

Seth Walk, Montana State University

Nick Pinkham, Montana State University


The impact of the microbiome on neuronal health is becoming more appreciated as microbiome dysbiosis is increasingly more apparent in neurodegenerative diseases. However, few current studies are taking into account the role that central metabolism plays in mediating cellular networks of the gut-brain axis, which are impacted by metabolism. We hypothesize that the gut-brain-liver axis is disrupted in neurodegenerative diseases, and particularly in the developmental and progressive neurodegenerative disease Familial dysautonomia (FD), which stems from a genetic mutation in the gene encoding the ELP1 protein. We postulate that the gut-brain dysbiosis observed in FD further triggers metabolic dysfunctions that result in gut bacteria and host metabolite imbalances, thereby exacerbating neuronal dysfunction.

Our approach utilizes next generation microbial DNA sequencing alongside 1H NMR metabolomics of FD patients and mouse models to identify molecular processes underlying gut-brain-liver communication deficits in FD. Application of systems-level analyses to FD disease pathology has the potential to enhance our knowledge of the cross talk between neuronal health, the gut microbiome, and metabolic homeostasis.

To this end, we have begun to characterize changes in the metabolome and gut microbiome in FD mouse models and FD patients. The metabolomics data obtained using 1H NMR on FD patient serum and stool samples reveal trends pointing to widespread changes in metabolite levels, including amino acids and nitrogenous base derivatives. Other metabolites appear to be consistently elevated in FD patients. In addition to metabolomics information, microbial DNA 16S sequencing reveals significant perturbation of the gut microbiome in FD patients.

Overall, our metabolic, gut microbiome, and gut histological analyses provide valuable markers that appear to associate with known clinical phenotypes of FD patients and FD disease severity. From these collective data, we have begun designing intervention experiments to establish cause and effect in FD associated gut-neuronal functions and how bacterial and host-derived metabolite levels potentially impact the gut-brain axis in FD.

"An Association Between Acetylcarnitine and TCA Cycle Intermediates in Septic Shock Non-Survivors Suggests Mitochondrial Dysfunction Influences Mortality" (138)


Marc McCann, University of Michigan (Primary Presenter)

Theodore Salvatore Jennaro, University of Michigan College of Pharmacy

Alan E. Jones, University of Mississippi Medical Center

Michael A Puskarich, University of Minnesota - School of Medicine

Kathleen A. Stringer, University of Michigan


Background: Sepsis induces a widespread disruption in metabolism. This is illustrated by the reported increases in acetylcarnitine (C2) and tricarboxylic acid (TCA) cycle intermediates present in sepsis non-survivors compared with survivors. C2 is thought to be produced via a shunt pathway during acetyl-CoA accumulation when entering the TCA cycle. Previous studies have suggested that sepsis non-survivors have higher levels of TCA cycle metabolites but their associations with C2 have not been studied in the context of the metabolic disruption induced by sepsis. Therefore, the purpose of this analysis was to investigate the relationship between C2 and various TCA cycle metabolites in patients with septic shock. We hypothesized there would be stronger associations between C2 and TCA metabolite levels in septic shock non-survivors than in survivors.

Methods: Serum samples from a clinical trial in septic shock patients were analyzed with targeted acylcarnitine (liquid chromatography-mass spectrometry) and TCA cycle metabolite (Ion-pairing chromatography MS) assays. The current analysis was conducted using pre-treatment (T0) serum samples of the placebo group patients (n=60). The metabolite concentrations were log transformed and scaled. Multiple linear regression models were employed to identify differences in the relationships between measured C2 and TCA cycle metabolite levels (citrate/isocitrate, fumarate, malate, succinate, 𝛼-ketoglutarate) in septic shock survivors and non-survivors at 28-days and 1-year. Herein, the models were developed using C2 as the dependent variable with 3 independent variable terms; a TCA cycle metabolite (TCA), mortality, and an interaction between TCA and mortality. They were constructed so that the coefficient of the interaction term would represent a difference in the TCA-C2 relationship between the survivors and non-survivors. Hypothesis testing (t-statistic) was used to determine if a significant difference (p-value ≤ 0.05) between the groups existed. Student’s t-tests were used to corroborate the reported findings of increased C2 and TCA cycle metabolite concentrations in the non-survivors. All statistics were conducted using Rstudio.

Results: A difference in the association between C2 and TCA cycle intermediates were found to be significant in 3 of the models that included 2 distinct TCA cycle metabolites (malate and succinate). The interaction term coefficient between malate and 28-day mortality was -0.528 (p= 0.05). For malate and 1-year mortality, the interaction term coefficient was -0.5461 (p= 0.03). The coefficient of the interaction between succinate and 1-year was -0.6624 (p= 0.04). Non-survivors had significantly elevated levels of malate (1-year: p= 0.02), succinate (1-year: p= 0.04), and C2 (28-days: p= 0.0009; 1-year: p= 0.002).

Conclusions: Elevations in the acylcarnitines, including C2, generally reflect impaired mitochondrial fatty acid oxidation. These results suggest C2 is representative of a more broadly encompassing mitochondrial dysfunction that includes the TCA cycle. The distinction between mortality groups also indicates that the disrupted metabolic pathways may influence survival.

Computational Approaches

"Semi-untargeted analysis of metabolic dynamics through network construction" (40)


YUE WU, Institute of Bioinformatics, University of Georgia, Athens, GA, USA (Primary Presenter)

Michael Thomas Judge, Department of Genetics, University of Georgia, Athens, GA, USA

Jonathan Arnold, Institute of Bioinformatics, University of Georgia, Athens, GA, USA; Department of Genetics, University of Georgia, Athens, GA, USA

Arthur S Edison, University of Georgia


Measurement and analysis of metabolic dynamics are crucial steps toward understanding biochemical adaptions and regulation. CIVM-NMR (continuous in vivo monitoring of metabolism by NMR) has enabled measurements of metabolic dynamics, and the software tool RTExtract (Ridge Tracking based Extract) recently provided a semi-automatic approach for efficient feature quantification when overlap and chemical shift changes are present. These approaches allow accurate, multi-dimensional, in vivo time series measurements with limited effort. For Neurospora crassa, NMR measurements were obtained under aerobic and anaerobic conditions, and ~300 features were extracted through time for each of six samples. Many features had interesting time dynamics and changed between conditions. These data provided a novel, detailed landscape of in vivo metabolism.

However, there are obvious difficulties in analyzing this data set with conventional “targeted” approaches in which features are annotated to compounds and interpreted within canonical biochemical pathways. First, feature annotation filters out most peaks, including many with interesting patterns. Second, existing knowledge of biochemical pathways is considerable but limited to specific functional states. For organisms like Neurospora crassa, existing metabolic maps are still incomplete and limit analysis of metabolic regulation under specific perturbations (different media and genetic mutants).

Instead of previous targeted analysis, we analyzed the data in a semi-untargeted way, through computing dependencies between all features (with or without annotations) and without restrictions of known pathways. This workflow is less targeted as no prior biochemical pathway was required and all features were analyzed. The targeted part comes from compound identification (ID). Dependencies among features were inferred by CausalKinetiX, where the data was processed by smoothing spline, and stable regression models were selected. These models describe stable kinetic dependencies between features and derivatives of features, providing cleaner signals for intermolecular biochemical relationships than correlation. The selected dependencies were summarized in a network, and community-based clustering (GLAY) was used to find consistent structures among all samples. When we applied limited compound ID information to these structures, we found that multiple functional clusters were uncovered, including one related to central energy metabolism that was supported by bootstrapping. Besides biochemical analysis, the dependencies were also transformed and combined with the correlation between features to obtain intramolecular associations. We produced separated clusters of intramolecular features and obtain more distinct clustering than just correlation. This approach can be a starting point for annotation.

A semi-untargeted approach was implemented for analyzing time series metabolomic measurements. Both biological interpretation and chemical annotation were facilitated. Further improvement can be done by collecting data with more diverse perturbations, where more functional states can possibly be discovered. This approach can be applied for time series metabolomic measurements from other technologies (e.g. mass spectrometry) and organisms.

"Differential Network Enrichment Analysis deciphers disease-related alterations in metabolism from high-throughput metabolomics data" (48)


Gayatri Iyer, University of Michigan - Ann Arbor (Primary Presenter)

Janis Wigginton, Michigan Regional Comprehensive Metabolomics Resource Core, Ann Arbor, MI

William L Duren, Department of Computational Medicine and Bioinformatics, University of Michigan Medical School, Ann Arbor, MI and Michigan Regional Comprehensive Metabolomics Resource Core, Ann Arbor, MI

Marci Brandenburg, Department of Computational Medicine and Bioinformatics, University of Michigan Medical School, Ann Arbor, MI and Taubman Health Sciences Library, University of Michigan Medical School, Ann Arbor, MI

Jennifer LaBarre, Department of Nutritional Sciences, University of Michigan School of Public Health, Ann Arbor MI

Charles Burant, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor MI

George Michailidis, Michigan Regional Comprehensive Metabolomics Resource Core, Ann Arbor, MI and Department of Statistics, University of Florida, Gainsville, FL

Alla Karnovsky, Department of Computational Medicine and Bioinformatics, University of Michigan Medical School, Ann Arbor, MI


We present a novel bioinformatics tool for the analysis and interpretation of Mass Spectrometry data. Based on our previously published DNEA (Differential Network Enrichment Analysis) algorithm, we have built a Java-based, user-friendly software that performs joint estimation of the partial correlation network topology from the input data, identifies subnetworks via consensus clustering, and determines enrichment using the NetGSA algorithm. We incorporated additional functionalities into the tool, allowing it to handle realistic sample sizes and experimental design in the data. In a high dimensionality setting when the number of metabolites is much larger than the number of samples in the input data, the DNEA tool uses a combined knowledge- and data-driven approach to aggregate highly correlated metabolites into singular features to make the feature space comparable to the sample space, hence overcoming the sparsity issue and achieving higher statistical power. Additionally, we use a subsampling-based procedure to recover highly robust edges in the network from datasets with extremely imbalanced experimental groups.

We applied the DNEA tool to analyze metabolomics data from Type 1 diabetic mice and humans with incident Type 2 Diabetes. Our analysis revealed enriched subnetworks related to oxidative stress, nucleotide and energy metabolism. We also identified enriched lipid subnetworks composed of Lysophosphatidylcholines and Lysophosphatidylethanolamines in pregnant mothers’ plasma that had a strong association with the birthweight of their infants, suggesting that subnetworks identified by DNEA are not only biologically relevant to the disease under study, but are also predictive of other phenotypes of interest. Such enriched subnetworks from partial correlations therefore have the potential to provide deeper insights into disease mechanisms as well as changes in the biochemistry under different physiological states.

"Lipid class prediction from MS/MS spectra using a NIST hybrid search-based machine learning workflow" (66)


Uri Keshet, UC Davis (Primary Presenter)

Tong Shen, NIH-West Coast Metabolomics Center, University of California, Davis

Raquel Cumeras, NIH-West Coast Metabolomics Center, Univeristy of California, Davis

Tobias Kind, UC Davis Genome Center - Metabolomics

Oliver Fiehn, UC Davis


With increasing sizes in experimental and virtual MS/MS libraries, compound annotation becomes routine. Yet, often only 10-30% of all MS/MS spectra are annotated per study, leaving thousands of peaks unused in biological reports. When multiple MS platforms are used with iterative MS/MS exclusion lists, combined metabolomic and lipidomic reports now commonly contain thousands of acquired MS/MS spectra. What can we tell about these compounds? Predicting chemical classes for all acquired MS/MS spectra could a) provide an extra layer of information to strengthen or contradict library annotations, b) help elucidate biochemical trends in the class level, rather than the individual metabolites and c) predict chemical class information on otherwise ‘known unknowns’ and thus at least render them useful for chemical set enrichment statistics and give hints on differences in regulation in biomedical studies.

We developed a workflow that starts with generating an array of MS/MS similarity scores using the NIST hybrid search function. Computing the correlations between these arrays that served as descriptors for machine learning prediction of compound classes. We demonstrate this workflow on a lipidomics dataset with 4,400 acquired MS/MS spectra in positive ionization mode using iterative exclusion lists on a Q-Exactive HF+ instrument (Thermo Fisher Scientific). MS-DIAL 3.96 was used for data analysis, deconvolution, retention time alignment, and library identification of the study, resulting in an initial 1,000 compound annotations. This annotated subset of the data was used for training and validation using a suite of machine learning models. The optimal model was then used to predict the full range of spectra in the study, including the remaining 3,400 MS/MS ‘known unknowns’. Next steps include using retention time and other experimental data to improve machine learning model accuracies.

"ADAP-KDB: Tracking and prioritizing unknown compounds across multiple studies with ADAP Spectral Knowledgebase." (120)


Aleksandr Smirnov, University of North Carolina at Charlotte (Primary Presenter)

Yunfei Liao, University of North Carolina at Charlotte

Eoin Fahy, University of California San Diego

Shankar Subramaniam, University of California San Diego

Xiuxia Du, University of North Carolina at Charlotte


The NIH's Metabolomics Data Repository (NMDR) has been growing steadily with data from an ever-increasing number of studies that use the liquid chromatography (LC-) and gas chromatography coupled to mass spectrometry (GC-MS) analytical platforms. Accompanying this growth is the enormous number of known and unknown compounds contained in that data. Tracking these compounds across multiple studies and making them searchable is an essential step towards an efficient sharing and aggregating the vast volume of data available in the metabolomics community. ADAP Spectral Knowledgebase (ADAP-KDB) has been developed to address this need. It uses raw data that is publicly available through NMDR to construct a library of consensus GC-MS and LC-MS/MS spectra of compounds observed in multiple studies. The mass spectra derived from studies in NMDR and used to construct the consensus spectra in ADAP-KDB can be traced back to NMDR with the URLs provided.

The constructed library of consensus spectra is available through ADAP-KDB online portal, where users can browse consensus spectra, upload their GC-MS or LC-MS/MS spectra, and match them to the library using the spectral similarity. Furthermore, users can filter the matching results based on the species, sample source, and disease information collected for each study. Finally, ADAP-KDB provides users with quantitative measures that could be used for prioritizing unknown mass spectra for subsequent compound identification. These quantitative measures include p-values of the ANOVA significance tests performed for each mass spectrum in every study, Gini-Simpson diversity index, and the Chi-Squared goodness-of-fit tests performed on distributions of the meta information collected for each study. This meta information includes information about species, sample source, treatment, MS instrument name and type, chromatography system, and column.

In order to build ADAP-KDB, 55 studies with raw untargeted GC-MS data were downloaded from NMDR. Their raw data was processed using the algorithms of ADAP-GC 4.0 workflow. As a result of the data processing, 20085 pure fragmentation mass spectra were constructed and then annotated by matching them against NIST EI Spectral Library with the matching score 800 or higher. Next, both annotated and unannotated spectra were uploaded to ADAP-KDB together with the meta information about each study. Then, all spectra from all studies were clustered based on their spectral similarity. Finally, a consensus spectrum was constructed for each cluster. The resulting 16878 consensus spectra are searchable and available through the online portal of ADAP-KDB.

Thus far, the following NMDR studies have been processed and the resulting spectra have been added to the ADAP-KDB: ST000025, ST000048, ST000054, ST000058, ST000061, ST000063, ST000065, ST000083, ST000084, ST000085, ST000087, ST000092, ST000117, ST000118, ST000189, ST000215, ST000302, ST000396, ST000417, ST000481, ST000495, ST000548, ST000603, ST000622, ST000658, ST000660, ST000832, ST000833, ST000878, ST000884, ST000889, ST000889, ST000910, ST000949, ST000950, ST000951, ST000952, ST000964, ST000965, ST000967, ST000968, ST000969, ST000979, ST000981, ST001034, ST001037, ST001044, ST001045, ST001046, ST001049, ST001050, ST001056, ST001075, ST001117, ST001190. Currently, we are working on processing studies with LC-MS/MS data from NMDR and expect to add LC-MS/MS spectra to ADAP-KDB in the coming months.

"SPECTRe: Substructure Processing, Enumeration, and Comparison Tool Resource: An efficient Python tool to encode all substructures of molecules represented in SMILES" (122)


Yasemin Yesiltepe, Washington State University (Primary Presenter)

Ryan Renslow, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA

Thomas Metz, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA


Substructures are chemical fingerprints of molecular structures, generally a significant moiety elucidating chemical or biological functionality, or a core element representing key chemical class information. Substructures are important molecular descriptors that contribute to the interpretation of chemical properties of biomolecules, and extensively used in many pharmaceutical applications, fragment-based drug discovery, de novo design of novel compounds, and identification of structural features.

One type of such fingerprints, structural keys-based fingerprints, mostly use 2D molecular graphs. Substructures are identified depending on the presence of the fingerprints/fragments in a supplied list of structural keys. For example, PubChem fingerprints have 881 keys used for similarity searching. They cover most of the chemical features for drug discovery and pharmaceutical science; however, the substructures are based on the presence or absence of the predefined keys in compounds and only useful when used with those molecules that are already likely composed of similar fragments.

Alternatively, path-based fingerprints address the lack of generality of the structural keys and are extensively used for fast substructure searching and filtering. They create chemical fingerprints by following paths and then hash every one of these paths into fixed-length bit vectors. However, it is possible to have an identical bit-string-fingerprint represented by two or more non-identical features due to hash collisions. Another limitation is that they can encode substructures up to a certain number of bits and may miss many substructures of large molecules. For example, the Daylight fingerprint encodes paths up to a given length consisting of up to 2048 bits. It is a computationally challenging problem to analyze branched and cyclic substructures. Hence, they provide only linear paths up to an adjusted number of bonds. For example, a path-based fingerprint, FP2, available in Openbabel indexes only linear segments of the maximum length of 7 atoms.

We introduce an open-source tool, SPECTRe, designed to provide all substructures/fingerprints of a given molecular structure. SPECTRe describes molecules in a 2D molecular graph where the nodes represent the atoms and the edges represent the bonds. A set of features of each atom (i.e. type, state of charge, hydrogen count, stereo configuration and valence) is attached to the corresponding vertex. Similarly, a set of features of each bond (i.e. bond type, vertex indices, and bond topology) is attached to each edge. An adjacency list is constructed using MOL files. Finally, we identify substructures via classical graph traversal (breadth-first and depth-first search) algorithms. No substructure is generated twice within a single run. Substructures are both stored in SDF files and represented in Canonical Smiles format.

We validated our tool with a set of 10,375 molecules, taken from HMDB, in the molecular weight range from 27 to 350 Da (<=26 non-hydrogen atoms), and spanning a wide array of structure-based chemical functionalities and chemical classes. Our results demonstrate the ability of SPECTRe to provide all substructures efficiently and shows promise to be used in many uses/areas of cheminformatics such as virtual screening for drug discovery, property prediction, fingerprint-based molecular similarity searching, data mining in finding of frequent substructures.

AutoCCS: a software tool to determine collision cross section from ion mobility spectrometry-mass spectrometry data" (146)


Joon-Yong Lee, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA (Primary Presenter)

Aivett Bilboa, Environmental Molecular Sciences Laboratory, Pacific Northwest National Laboratory, Richland, WA

Chris Conant, Pacific Northwest National Laboratory

Kent Bloodsworth, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA

Daniel Orton, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA

Ian Webb, Department of Chemistry & Chemical Biology, Indiana University–Purdue University Indianapolis, Indianapolis, IN, 46202, USA

Kim Hixson, Earth and Biological Sciences Directorate, Pacific Northwest National Laboratory, Richland, WA, 99352, USA

John Fjeldsted, Agilent Technologies, Santa Clara, CA, USA

Yehia Ibrahim, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA

Samuel Payne, Department of Biology, Brigham Young University, Provo, UT, 84602, USA

Erin Baker, North Carolina State University

Christer Jansson, Earth and Biological Sciences Directorate, Pacific Northwest National Laboratory, Richland, WA, 99352, USA

Richard Smith, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA

Thomas Metz, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA


Metabolomics is a rapidly growing field of study involving the quantitative analysis of small molecules resulting from metabolic processes in biological systems. While typically these studies are performed using gas or liquid chromatography-mass spectrometry (MS) systems, ion mobility spectrometry (IMS) coupled with MS is increasingly utilized in metabolomics as it offers high resolution and rapid separations (milliseconds to seconds). IMS-MS separates ionized molecules based on their structural characteristics such as size, shape, and charge state and enables characterization of ions by their drift times. The measured drift time for each molecule can be translated to the collision cross section (CCS), a chemical-physical molecular property that represents the three-dimensional structure of the corresponding ion. IMS-MS offers high throughput two-dimensional separations that increase specificity in molecular identifications and, e.g., via separations of isobars and isomers. Recently, the repeatability and reproducibility of CCS measurements has been evaluated for a wide range of molecules across different instruments and laboratories achieving highly consistent results. However, an open-source computational tool to automatically calculate CCS values for analyte ions measured in both targeted and untargeted IMS-MS analyses is lacking, and manual processing with instrument vendor software is still required. Therefore, data processing in large-scale IMS-MS studies is a bottleneck.

Here, we present an open-source, command-line software tool, AutoCCS, that allows users to automatically determine CCS values from various IMS-MS measurements. AutoCCS was written in Python for greater accessibility and is independent from other data processing layers, e.g., IMS feature finding. The main input for AutoCCS is a list of IMS-MS features, i.e., a list of mass-to-charge ratios and drift times of analyte ions. AutoCCS provides CCS calculation in several formats: 1) stepped field methods for drift tube-based IMS (DTIMS), 2) single field methods for DTIMS with linear calibration, and 3) non-linear calibration methods for traveling wave based-IMS (TWIMS), e.g., Structures for Lossless Ion Manipulations (SLIM). For single-field methods, AutoCCS supports internal or external calibration (with a list of reference ions) and an enhanced CCS calibration mode with correction for temperature and pressure variations that may occur during large-scale studies. For CCS determination in TWIMS experiments, AutoCCS allows either a polynomial or power regression analysis to compute a calibration function. AutoCCS generates comprehensive and informative tables and figures for users to quickly evaluate the quality of CCS determination. To demonstrate the effectiveness and utility of AutoCCS for CCS calculation, we conducted various IMS experiments and AutoCCS accurately and reproducibly calculated CCS values of diverse ions analyzed in various IMS experiments even with different IMS instrumentation.

"CCSP 2.0: An Open Source Jupyter Tool for the Prediction of Ion Mobility Collisional Cross Sections in Metabolomics." (156)


Markace Alan Rainey, Georgia Institute of Technology (Primary Presenter)

Facundo M. Fernández, School of Chemistry and Biochemistry, Georgia Institute of Technology

Chandler Watson, Georgia Institute of Technology


The definitive chemical identification of metabolites is an enduring challenge for untargeted metabolomics studies. Ion-mobility (IM) can aid in this endeavor by offering an additional dimension for the rapid separation of complex mixtures that is orthogonal to liquid chromatography. IM enables the measurement of collisional cross section (CCS) values for each molecule. This CCS value is more reproducible between laboratories and instrumental platforms than chromatographic retention times and can be combined with high resolution mass spectra and nuclear magnetic resonance spectroscopy experiments to facilitate metabolite annotation. Recent work has demonstrated that CCS values can be accurately predicted using machine learning (ML) approaches.1-3 The benefit of these efforts has not yet extended to a wider metabolomics audience, however. The use of proprietary software and highly specialized computational techniques has hindered the routine inclusion of CCS prediction in metabolite identification workflows. Here, we present the CCS Predictor 2.0 (CCSP 2.0), an open-source tool that can be easily used by researchers to predict CCS values of metabolite candidates. The tool utilizes the Python package Mordred to calculate 1613 2D molecular descriptors for use in partial least squares regression (PLSR) modeling. CCSP 2.0 is packaged as a Jupyter Notebook, which allows the seamless integration of text and coding blocks. This format allows novice users to follow the coding logic and allows more experienced users the flexibility to customize full workflows with Jupyter. Because CCSP 2.0 provides users with the flexibility to choose their own training sets, machine learning models can be fine-tuned to predict the CCS values of custom chemical datasets. Assessments of the accuracy of CCSP 2.0 using a chemically diverse subset of compounds (n=1311 ) in the McLean Unified CCS Compendium resulted in R2 values above 0.98 for cross-validated calibration data with median CCS prediction errors below 2.3% and a root mean square error (RMSE) below 9.0 Å2 . When calibrated with chemically similar compounds, CCSP 2.0 performance improves. For example, assessment with the lipid and lipid-like molecules superclass of the McLean Compendium (n=99) yielded a median prediction error of 0.70% and a RMSE of 7.2 Å2 . Notably, the entire computational process can be completed within five minutes on a laptop computer with average specifications. This streamlined, open-source tool also remains customizable to the user, allowing for the implementation of additional pre-processing and data analysis methods as needed.

1. Plante, P.-L.; Francovic-Fontaine, É.; May, J. C.; McLean, J. A.; Baker, E. S.; Laviolette, F.; Marchand, M.; Corbeil, J., Predicting Ion Mobility Collision Cross-Sections Using a Deep Neural Network: DeepCCS. Analytical Chemistry 2019, 91 (8), 5191-5199.

2. Soper-Hopper, M. T.; Petrov, A. S.; Howard, J. N.; Yu, S. S.; Forsythe, J. G.; Grover, M. A.; Fernández, F. M., Collision cross section predictions using 2-dimensional molecular descriptors. Chemical Communications 2017, 53 (54), 7624-7627.

3. Zhou, Z.; Shen, X.; Tu, J.; Zhu, Z.-J., Large-Scale Prediction of Collision Cross-Section Values for Metabolites in Ion Mobility-Mass Spectrometry. Analytical Chemistry 2016, 88 (22), 11084-11091.

Ecology and Environment

"Metabolomic Signatures of Coral Bleaching History" (52)


Robert Quinn, Department of Biochemistry and Molecular Biology (Primary Presenter)

Christian Martin H., 1 Department of Biochemistry and Molecular Biology. Michigan State University

Ty Roach, Hawaii Institute of Marine Biology

Ford Drury, Hawaii Institute of Marine Biology

Arthur Daniel Jones, Michigan State University

Jenna Dilworth, Hawaii Institute of Marine Biology


Coral bleaching, a process where corals expel their photosynthetic symbionts, has a profound impact on the health and function of coral reefs. As global ocean temperatures continue to rise, bleaching poses the greatest threat to coral reef ecosystems. Here, untargeted metabolomics was used to analyze the biochemicals in pairs of adjacent corals from a patch reef in Kāneʻohe Bay, Hawaiʻi where one colony in the pair bleached (in 2015) and recovered while the other did not bleach. There was a strong metabolomic signature of prior bleaching history four years after recovery found in both the host and its algal symbionts. Machine learning analysis determined that the strongest metabolite drivers of the difference in bleaching phenotype were a group of betaine lipids. Those with saturated fatty acids were significantly enriched in thermally tolerant corals and those with longer, unsaturated and diacyl forms were enriched in historically bleached corals. Host immune response molecules, Lyso-PAF and PAF, were also altered by bleaching history and were strongly correlated with symbiont community and algal-derived metabolites suggesting a role of coral immune modulation in symbiont choice and bleaching response. To validate these findings, we tested a separate in situ set of corals and were able to predict the bleaching phenotype with 100% accuracy. Furthermore, corals subjected to an experimental temperature stress had strong phenotype-specific responses in all components of the holobiont, which served to further increase the differences between historical bleaching phenotypes. Thus, we show that natural bleaching susceptibility is simultaneously manifested in the biochemistry of the coral animal and the algal symbiont and that this bleaching history results in different physiological responses to temperature stress. This work provides insight into the biochemical mechanisms involved in coral bleaching and presents metabolomics as a valuable new tool for resilience-based reef restoration.

"Assessing soil organic matter features as detected with direct infusion high resolution mass spectrometry and LC-MS-MS feature based molecular networking" (118)


Nicole DiDonato, Pacific Northwest National Laboratory (Primary Presenter)

Chaevien Clendinen, Pacific Northwest National Laboraotry

Albert Rivas-Ubach, Pacific Northwest National Laboratory

Nikola Tolic, Pacific Northwest National Laboratory

Noah Sokol, Lawrence Livermore National Laboratory

Dinesh Adhikari, Lawrence Livermore National Laboratory

Carmen Enid Martínez, Cornell University

Jennifer Pett-Ridge, Lawrence Livermore National Laboratory

Ljiljana Pasa-Tolic, Pacific Northwest National Laboratory


Molecular formula assignment using direct infusion high resolution mass spectrometry has become a powerful tool for untargeted analysis of natural organic matter. However, these measurements are limited by charge competition, a lack of metabolite identification and the unknown diversity of isomeric structures. Although complete separation of natural organic matter is not yet achievable, chromatography can reduce charge competition, allow for separation of isomers and LC-MS/MS measurements enable molecular identification. The lack of chromatographic separability of organic matter is suggested as evidence for a large number of isomers per assigned formula and ubiquitous fragmentation patterns indicate a high degree of isomeric structural similarity2. However, much lower numbers of isomers have been estimated based on detection of hundreds of chromatographically separated features3. Few studies have attempted to identify metabolites from soils, of which only a fraction can be annotated due to lack of comprehensive libraries4,5.

In this work, LC-MS/MS feature based molecular networking6 is investigated and compared with direct infusion features to corroborate molecular formula assignments and provide additional structural information. The fraction of formulas observed in direct infusion that can be separated, matched with existing libraries, categorized by neutral losses or as isomers is also assessed.

Preliminary results suggest LC-MS may be able to separate a significant portion of direct infusion features, however not all of these features are confidently assigned to molecular formulas. By comparing untargeted data-dependent LC-MS/MS features from 39 samples and standards, we found 5-30% of direct infusion assigned formulas per sample could be networked using global natural product social (GNPS) feature based molecular networking. We found good agreement between GNPS annotations and direct infusion molecular formula assignments. Unique annotations include polyphenols, lipids, terpenoids and carboxylic acids, expected components of natural organic matter. Although multiple orthogonal chromatographic separations may be required for a comprehensive evaluation of the structures and diversity of isomers, LC-MS-MS feature based molecular networking can complement direct infusion measurements by validating direct infusion formula assignments, improving assignment rates and inferring additional structural information. This, combined with the increased number of features and potential for networking similar fragmentation patterns of unknowns into specific classes of compounds are the next steps to understanding the molecular structures of carbon that persist in various environments.

[1] Tfaily, M. M., et al. (2013). Geochim et Cosmochim Acta 112: 116-129.

[2] Leyva, D., et al. (2019). Faraday Discuss 218(0): 431-440.

[3] Lu, K. and Z. Liu (2019). Front Mar Sci 6(673).

[4] Ladd, M. P., et al. (2019). Sci Rep 9(1): 5810.

[5] Nguyen, T. D., et al. (2020). Metabolites 10(3): 86.

[6] Wang, M et al. (2016) Nat Biotechnol 34.8: 828-837

Flux Profiling

"An untargeted LC-MS workflow to measure stable isotope incorporation from acquisition to analysis" (142)


Amanda Souza, Thermo Fisher Scientific (Primary Presenter)

Ioanna Ntai, Thermo Fisher Scientific

Tatjana Talamantes, Thermo Fisher Scientific

Ralf Tautenhahn, Thermo Fisher Scientific


Stable isotope labeling (SIL) in metabolomics is a notable tool to elucidate metabolic pathway associations (tracer analysis) and rate of change of metabolites (flux analysis). Analytical methods employing SIL typically focus on defined pathways of interest employing a targeted approach, yet the interconnected network of metabolic pathways may involve unappreciated pathways. Further, SIL is used to increase confidence in unknown identification and to determine true sample related mass spectral features in untargeted analysis. As the role of SIL in untargeted metabolomics expands, the development of a facile, integrated workflow to support data acquisition and processing is needed. Here we demonstrate an untargeted metabolomics approach applied to 13C-labeled E. coli extracts using intelligent data-dependent acquisition and a single processing software platform.

Dried down E. coli cell extracts purchased from Cambridge Isotope Laboratories were reconstituted in water containing 0.1% formic acid. Unlabeled and 13C-labeled extracts were combined at defined ratios to obtain samples including 0%, 10%, 25%, 50% of 13C-labeled metabolites. Metabolite separation was achieved with a Thermo Scientific™ Hypersil GOLD™ column (15cm x 2.1mm ID) on a Thermo Scientific™ Vanquish™ UHPLC system coupled to a Thermo Scientific™ Orbitrap Exploris™ 240 mass spectrometer. Full scan data were collected in positive mode using 120K resolution while fragmentation spectra were collected using Thermo Scientific™ AcquireX intelligent data acquisition. Data were processed using Thermo Scientific™ Compound Discoverer™ software for SIL incorporation analysis, unknown annotation, and pathway mapping.

Untargeted LC-MS data acquisition combined with untargeted data processing was applied to an SIL experiment with the aim of analyzing global incorporation levels instead of focusing on select targets. Confident ion assignments were achieved with accurate mass measurements and higher resolution settings to distinguish isotope pattern, isotope fine structure and isotopologues with slightly different m/z.

Confident annotation assignments of all metabolites were carried out using a consensus approach from multiple annotation sources including elemental composition prediction, MS1 database searching against the ChemSpider™ database, MS/MS spectral matching against the Thermo Scientific™ mzCloud™ library and an in-house spectral library generated from commercially available standards.

To assess the level of 13C-label present in mixtures containing different ratios, presence of isotopologues for annotated compounds were compared to unlabeled extract. Excellent isotopic fidelity and enhanced fine structure at higher resolution, enabled accurate quantitation of 10% incorporation of 13C label. Incorporation rates were detected at expected levels corresponding to defined ratios of unlabeled to labeled for key pathway metabolites like amino acids, modified amino acids and nitrogenous bases. Reproducibility of isotopologue determination at different levels of 13C incorporation generated low %CV vales for all ratio mixtures.

Applying this approach, several unexpected putative metabolites like tyrosine methyl ester were detected with incorporated 13C-label. Putative metabolites with label incorporation were further interrogated for pathway association using the Thermo Scientific™ Metabolika™ biological pathway database to helped discern metabolic activity.

Unlike targeted SIL analysis, where a hypothesis is needed before designing the experiment, this newly developed SIL workflow allows for detection of incorporation without biological knowledge a priori and can be used as a discovery tool.

"Comparing gas chromatography with time of flight, quadrupole time of flight and quadrupole mass spectrometry for stable isotope tracing" (176)


Ying Zhang, UCDavis (Primary Presenter)

Bei Gao, UCSD

Oliver Fiehn, UC Davis


Stable isotope tracers are applied in vivo and in vitro studies to reveal the activity of enzymes and intracellular metabolic path-ways. Most often, such tracers are used with gas chromatography and mass spectrometry (GC-MS) due to its ease of operation and reproducible mass spectral databases. Differences in isotope tracer performance of classic GC-quadrupole MS instruments to new-er time-of flight instruments are not well-studied. Here, we used of three commercially available instruments on identical samples of a stable isotope labeling study that used U-13C6 glucose to investigate the metabolism of Rothia mucilaginose with respect to 29 amino acid and hydroxyl acids involved in primary metabolism. Overall, all three GC-MS instruments were capable of performing stable isotope tracing studies for glycolytic intermediates, TCA metabolites and amino acids, yielding similar biological results.


"LipidQuan: An Interlaboratory Evaluation of Rapid Lipid Quantification" (24)


Nyasha Munjoma, Waters Corporation, Wilmslow, UK (Primary Presenter)

Giorgis Isaac, Waters Corporation, Milford, MA

Nicola Gray, Murdoch University

J. Will Thompson, Duke University School of Medicine

Samantha Ferries, Waters Corporation

Stephen Wong, Waters Corporation

Anthony Midey, Waters

Steven Lai, Waters Corporation

Lee Gethings, Waters Corporation

Arthur Moseley, Duke University School of Medicine

Jeremy Nicholson, Murdoch University

Robert Stephen Plumb, Waters Corporation, Milford, MA, USA


Lipid compositions are tightly regulated through complex mechanisms. A range of disorders such as cancer and cardiovascular disease result in changes of the lipidome. Lipidomics is well placed to aid biological interpretation through more precise and accurate lipid analysis. A major challenge hampering the field is the large disparity in methodologies and technologies, resulting in discrepancies of published data and broader issues of irreproducibility. As the field of lipidomics progresses, it will be increasing critical to be able to control, minimize, or, at least acquire a better understanding of inter and intra laboratory variability in order to make more informed and accurate decisions. Here, we highlight a quantitative, targeted LC-MS workflow showing excellent inter and intra reproducibility across multiple laboratories.

We performed a multi-site validation study (consisting of seven laboratories) using a semi quantitative, LC-MS approach based on HILIC chromatography for screening lipids in plasma. Samples were extracted using an IPA-based protein crash procedure and analysed using a high throughput HILIC based gradient (8 minutes) coupled with a tandem quadrupole. Data were collected in positive, negative and positive/negative switching modes. Validation tests included precision, accuracy, linear response, LOQ, carryover, interference, dilution integrity and sample volume limit tests. The NIST SRM 1950 reference plasma was quantified using commercially available standards as internal standards and calibrants. Data from all laboratories were processed using TargetLynx (Waters) and Skyline (MacCoss Laboratory) to provide quantitative outputs.

The study protocol implemented by all laboratories was designed to meet the minimum FDA and EMA guidelines. Lipids across 14 different lipid classes were assessed with coefficient of variance (CV) and a mean bias of less than 25% being deemed as acceptable for the study. The retention time CV’s across a total of 42 QC injections was less than 1% for accepted data. The level of carryover for the analyte was considered as acceptable, if the response in the double blank sample was ≤20% of the average response from the acceptable LLOQ standards in the batch. The limit of carryover for internal standards was set to ≤5%, with 14 out of the 20 internal standards meeting these criteria. A minimum sample volume of 15 µL of plasma was found to produce acceptable results. Dilution integrity tests show samples can be diluted up to 5 times to ensure peaks are within the linear range for more accurate quantification. The results from all participating laboratories provided benchmarking of the LipidQuan workflow for intra- and interlaboratory quality control, demonstrating easy deployment and excellent reproducibility of the methodology.

"Lipidomics signatures of brain from patients with Alzheimer’s disease reveals a specific enrichment in glycerolipids, glycerophospholipids, and sphingolipids" (58)


Sumeyya Akyol, Beaumont Health System-Research Institute (Primary Presenter)

Zafer Ugur, Beaumont Health System-Research Institute

Ilyas Ustun, DePaul University

Ali Yilmaz, Beaumont Research Institute; Oakland University-William Beaumont School of Medicine

Santosh Kapil Kumar Gorti, SCIEX

Kyung Joon Oh, Beaumont Health System-Research Institute

Bernadette McGuinness, Centre for Public Health, School of Medicine, Dentistry and Biomedical Sciences, Queen’s University Belfast

Peter Passmore, Centre for Public Health, School of Medicine, Dentistry and Biomedical Sciences, Queen’s University Belfast

Patrick G. Kehoe, Dementia Research Group, Translational Health Sciences, Bristol Medical School, University of Bristol

Michael Maddens, Oakland University-William Beaumont School of Medicine

Brian D. Green, Institute for Global Food Security, School of Biological Sciences, Queen’s University Belfast

Stewart F. Graham, Beaumont Research Institute; Oakland University-William Beaumont School of Medicine


Introduction: Alzheimer`s disease (AD) is the most common manifestation of degenerative and progressive dementia among older adults and its etiopathophysiology has been reported to be closely linked with abnormal lipid metabolism.

Objectives: To gain a more comprehensive understanding of what causes AD and its subsequent development, we aim to profile the lipidome of postmortem (PM) human brain (neocortex) from people with a range of AD pathology.

Methods: Using high resolution mass spectrometry, we employed a semi-targeted, fully quantitative lipidomics profiling method called the Lipidyzer to biochemically profile brain from people with mild-AD (MAD; n=15), severe-AD (AD; n=16) and compared them with controls (n=16).

Results: We accurately identified and quantified 1143 lipid metabolites in PM brain extracts. Univariate analysis revealed that 420 lipids were at statistically (p<0.05; q<0.05) significantly different concentrations between AD and control, 49 lipids between MAD and control, 439 lipids between AD and MAD. We also highlight 13 different subclasses of lipids which we underline as being significantly perturbed. These include neutral lipids, glycerolipids, glycerophospholipids, and sphingolipids. Within these subclasses of lipids, most prominently perturbed individual lipids between AD and control were DAG(14:0/14:0), TAG(58:10/FA20:5), TAG(48:4/FA18:3), between MAD and control were PE(P-18:0/18:1), PS(18:1/18:2), PS(14:0/22:6), between AD and MAD were PE(P-18:0/18:1), DAG(14:0/14:0), PS(18:1/20:4).

Conclusion: We report the most comprehensive lipid profiling of PM brain from people suffering from AD and highlight a number of lipid metabolic pathways as being significantly perturbed due to the presence of the disease. Our study demonstrates the great potential of lipidomics for studying AD etiopathogenesis and for identifying potential early diagnostic biomarkers. In conclusion, previous findings and our current results provide further evidence for the role of altered lipid metabolism in the pathogenesis of AD.

"Triboelectric Nanogenerator Ion Mobility-Mass Spectrometry for the Comprehensive Structural Elucidation of Unsaturated Glycerophospholipids." (80)


Marcos Bouza Areces, Georgia Tech (Primary Presenter)

Yafeng Li, Georgia Institute of Technology

Zhong L. Wang, Georgia Institute of Technology

Facundo M. Fernández, School of Chemistry and Biochemistry, Georgia Institute of Technology


Static electricity is one of the oldest means for generating electrical charge and one of the least explored phenomena in metabolomics and mass spectrometry (MS). In previous reports we have shown that triboelectricity can be harnessed through triboelectric nanogenerators (TENG) for producing ions (Nat. Nanotechnol. 2017, 12, 481-487) for MS analysis, and how TENG with larger electrodes operate in a dual ESI/APCI regime (JASMS, 2020, 31, 727-734). Here, we harnessed the gas-phase oxidation reactions observed during TENG-MS for double bond pinpointing in complex glycerolipids. By coupling TENG to time-aligned parallel (TAP) analysis in an ion mobility (IM) mass spectrometer, comprehensive structural information for lipids was produced. Depending on the ionization polarity used, this approach allowed double bond position pinpointing and/or sn-fatty acid chain position localization. Overall, TENG-TAP IM-MS is seen as a rapid, simple and robust technique for shotgun lipidomics without the need of ozonolysis, ultraviolet photodissociation, or Paternò–Büchi reagents, as used in previous approaches.

To perform TENG MS experiments, glass nanoelectrospray emitters (Econo 12, New Objective) were powered using a large-area sliding freestanding (SF) TENG. A Synapt G2S (Waters) IM-mass spectrometer operated in TAP mode was used for all experiments. The sliding layer of the TENG device was made of Nylon (12x12cm), whereas the stationary layer was made of fluorinated ethylene propylene (FEP, 24x12cm). A linear motor (Linmot) was used to actuate TENG reproducibility. Different organic solvents (methanol, acetone, or acetonitrile) combined with water and different additives (NH4OAc or LiOAc) were tested to optimize the yield of epoxidized mono and polyunsaturated glycerophospholipids.

The transient glow corona discharge burst observed during large-area TENG cycles enabled gas-phase reactions in the gap between the ion source and the mass spectrometer inlet, causing the formation of radical cations of non-polar species or gas-phase oxidation, not typically observable by nanoelectrospray alone. This rapid epoxidation reaction was leveraged for the analysis of unsaturated compounds, specifically unsaturated lipids. For testing purposes, a 20 µM solution of 1, 2-dioleyl-sn-glycero-3-phosphocholine (PC 18:1(Δ9Cis)) was used for TENG-TAP IM-MS experiments in positive and negative ion mode. In negative ion mode, the epoxide acetate adducts at m/z 860.6 ([MO+OAc]-) yielded the m/z 171.1 and 155.1 diagnostic fragments after pseudo MS3 TAP IM-MS experiments, indicating a double bond in the 9th carbon of the fatty acyl chains. For a single TENG pulse (<1s) the epoxidation yield observed was about 5%. Positive ion mode experiments were less efficient in terms of epoxidation yield, but more informative in terms of lipid structure elucidation. In this mode, the five-member ring dioxolane ion formed during TENG-TAP IM-MS enabled both double bond position pinpointing and sn-position elucidation; fragments observed at m/z 499.4 and 483.4 matched with the double bond location at the 9th carbon, whereas fragments at m/z 361.3, 345.3, 329.3, 303.3 and 287.3 revealed the sn-positions of the fatty acyl chains. This methodology was tested on different classes of glycerophospholipids with various degrees of unsaturation, double bond positions or sn compositions, as well as for biological extracts and real sample analysis with excellent results.

"Profiling the Lipidome: Quantitate up to 2000 Lipid Molecular Species in a Single Injection" (96)


Mackenzie J Pearson, Sciex (Primary Presenter)

Paul Norris, Sciex

Paul RS Baker, SCIEX

Santosh Kapil Kumar Gorti, SCIEX


The complexity of the lipidome is very high with a huge diversity in lipid molecular species making comprehensive quantitative profiling challenging. Retention time scheduling for MRM transitions during targeted assays enables more compounds to be quantified with higher quality results. In this assay, MRMs to quantitate almost 2000 lipid molecular species were combined into a single assay on the QTRAP® 7500. Amino column chemistry was chosen for lipid class separation to minimize isomeric interference. The target list of lipids is comprehensive, covering most major lipid classes and categories, and MRMs were selected to cover lipids containing fatty acids with 14 to 26 carbons and 0 to 6 double bonds. The method is customizable, so new lipid categories, classes and molecular species can be added to the MRMs list and is matrix agnostic. The sMRM Pro Builder template, which was developed to streamline the method optimization process, enables assigning the retention time, optimize dwell weight and set window size per MRM to enhanced coverage and sensitivity of the method. This optimization improved results quality especially on low abundant lipids. Lipid standards from 20 different classes, with either heavy isotopic labeled lipids or odd chain lipids (Lipidyzer Standards or SPLASH mix), served as internal standard. This method provided extensive lipid class coverages including, CE, CER, DCER, HCER, LCER, TAG, DAG, MAG, CL, LPC, PC, LPE, PE, LPG, PG, LPI, PI, LPS and PS.

Ten replicates of NIST plasma was used to test the assay. Compared to previously established methods, over 20% more lipids could be identified in a third less of the time. In a total 17-minute chromatographic run, 1046 lipid molecular species were quantified on the QTRAP® 7500. By utilizing the sMRM Pro Builder template, you can also automatically reject low intensity peaks (uncertain detection) from the final assay to further improve results quality by reducing MRM concurrency and improving sampling efficiency. This filtering step provided less than 20% CVs on over 80% of lipids quantified.

"Infusion MS/MSALL with Differential Mobility Separation : A High-throughput Lipidomic Solution for Untargeted Profiling" (112)


Mackenzie J Pearson, Sciex

Santosh Kapil Kumar Gorti, SCIEX

Paul Norris, Sciex

Darren Dumlao, SCIEX

Paul RS Baker, SCIEX (Primary Presenter)


Introduction (108/120 words) Untargeted lipidomic profiling is an advancing field with growing interest. Liquid chromatography has been the preferred method in this growing field to profile lipids, but there is no one LC method that effectively provides an exact precursor mass paired with fragment ion data to effectively profile most lipid molecular species. Infusion-based lipidomics, or Shotgun Lipidomics, has become an alternative approach. This method creates a detailed lipid profile across a specified mass range while also generating tandem mass spectrometry (MS/MS) data. This approach partnered with differential ion mobility separation to acquire lipid class specific data, in under twelve minutes, provides a high-throughput solution to measure lipid changes in biological samples. Methods 120 words Data was collected on a series of lipid samples using the Infusion MS/MSALL method on the TripleTOF® 6600+ system. The MSMSALL method on this system generates data by collecting product ions of all precursor masses, stepping across the precursor mass range in 1Da increments, collecting full scan MS/MS. The MS was also equipped with SelexION® Technology, to perform differential mobility separation (DMS) to isolate key lipid classes in the gas phase, before MS analysis. Chemical modifiers were used to adjust the lipid separation, based on their dipole moment. LipidViewTM software is used to analyze the data collected from the SelexION® runs in a high throughput fashion. Preliminary Data (377 words) Global lipid profiling is a primary means to measure the changes in lipid molecular species as a function of human disease or drug treatment. One of the most challenging aspects of lipid analysis by mass spectrometry is isobaric overlap, or molecules having the same or nearly the same mass. With direct infusion, thousands of lipids can be rapidly profiled in biological samples, and allows for retrospective analysis of acquired data. However the issue of isobaric overlap is not addressed. Isobaric overlap is a complex problem in lipidomic workflows and can make positive identification or quantification of a lipid species very difficult. The issue of isobaric overlap is often approached with liquid chromatography separation. However, lengthy sample preparation, method development, and lengthy chromatographic run times for adequate separation can diminish throughput. Differential ion mobility separation (DMS) uses a molecule’s dipole moment to isolate it rather than shape and size. The SelexION® device is installed at atmosphere, so the addition of chemical modifiers can be used to assist in additional separation. In this study, 1-propanol is used as the chemical modifier to separate the six phospholipid subclasses while cardiolipin’s -2H charge state is used to separate this class without the assistance of chemical modifier. Unique compensation voltages are then found with standards and assigned to each lipid class to ensure lipid class isolation. In negative ion mode, all seven classes can be acquired in just under twelve minutes of total run time. This allows for untargeted, high-throughput acquisition of the lipidome while simultaneously acquiring MS/MS spectra for lipid species identification. All samples from each class can be loaded in LipidViewTM software and a lipid class can be assigned along with its

"New annotation tools for advanced 4D-Lipidomics workflows." (116)


Sven Meyer, Bruker Daltonics

Ansgar Korf, Bruker Daltonics

Heino Martin Heyman, Bruker Scientific (Primary Presenter)

Aiko Barsch, Bruker Daltonics

Florian Zubeil, Bruker Daltonics


The annotation of lipids can be very demanding due to the high number of structural variations. Mass spectrometry-based identification typically relies on characteristic fragments from lipid headgroups and side chains obtained from MS/MS experiments. Depending on the quality of the MS/MS spectra, the depth of the structure elucidation can cover different levels like molecular formula level, chain composition level, etc. While the matching of MS/MS spectral libraries such as e.g. LipidBlast gives a broad and quick overview on the lipid content, the detail of the annotation level can be too high. The new tools applied in this study avoid this risk of over annotation and simplify the automatic identification and validation of lipid features by making use of selected fragmentation rules and by enabling visual investigation of retention time and Collisional Cross Section consistency within lipid groups.

For this study, NIST SRM 1950 plasma MTBE lipid extract was investigated using mobility-enhanced LC-MS/MS data acquired on a timsTOF Pro (Bruker Daltonics) in PASEF acquisition mode. The raw data were processed with a prerelease version of MetaboScape 2021 (Bruker Daltonics) using four-dimensional feature extraction and ion mobility-enhanced compound identification. The MS/MS spectra were annotated using a new rule-based annotation tool implemented in MetaboScape. The current version of this tool used [M+H]+, [M+Na]+, [M+NH4]+, [M-H2O+H]+, [M-H]-, [M+HCOO]- and [M+CH3OO]- ions from 28 sub-classes out of four main categories (Glycerolipids, Glycerophospholipids, Sphingolipids and Sterol lipids). The annotations were checked for consistency or possible outliers using a 4-dimensional Kendrick mass defect plot.

LC-TIMS-MS/MS data of the NIST SRM 1950 plasma lipid extracts were acquired as triplicate injections in positive and negative PASEF mode. All four-dimensional data were automatically processed with MetaboScape using the T-ReX 4D algorithm and the retention time aligned features listed in a bucket table. In the process of the feature extraction, all important qualifiers like exact mass, isotopic pattern quality, retention times, MS/MS spectra and CCS values were extracted for all specified adducts and neutral losses. To increase the data depth for lipid assignment, data of both polarities were merged. The novel rule-based annotation was applied to the complete feature list and the identifications were compared to annotations generated by MS/MS library matching using the open-source LipidBlast library. The rule-based tool showed a more conservative lipid assignment in accordance to the Lipidomics Standards Initiative guidelines compared to the library-based approach. A manual inspection of several examples revealed that the library matching over-annotated several lipids.

Furthermore, a four dimensional, CCS and retention time aware Kendrick mass defect plot was used to investigate several lipid classes for further non-annotated candidates in the series and to check for possible false annotations. Finally, the confidence of annotations was evaluated using the visual annotation quality scoring tool applied in MetaboScape to rate all qualifiers.

"Sensitive and Comprehensive Lipid Mediator Profiling using Advanced Scheduled MRM with Polarity Switching and QTRAP Enhanced Product Ion Scanning" (124)


Paul Norris, Sciex (Primary Presenter)

Santosh Kapil Kumar Gorti, SCIEX

Mackenzie J Pearson, Sciex


Lipid mediators regulate diverse physiological processes and play critical roles in modulating human inflammation-resolution responses. Understanding the complex roles and regulation of lipid mediators in human health relies on sensitive and comprehensive methods to capture complete pathway profiles. In biological samples, these bioactive lipids are often present at or below nanomolar concentrations and require a highly sensitive LC-MS/MS platform for identification and quantitation. In this work, a quantitative and qualitative workflow is described for the analysis of different lipid mediator species including specialized pro-resolving mediators (SPM), leukotrienes, prostaglandins, hydroxy-eicosatetraenoic acids (HETEs) and epoxy-eicosatrienoic acids (EETs), as well as metabolites and markers of reactive oxygen and nitrogen species. Using a QTRAP® 7500 system coupled with an ExionLC™ System, we developed a targeted panel of 88 compounds for comprehensive profiling of lipid mediators and pathway markers. A Kinetex® Polar C18 column was used for LC based separation of lipid mediators, epimers and other isoelemental structures within a 20 min run time. Advanced scheduled sMRM was used to optimize scanning windows and dwell weighting. The negative mode lipid mediator panel includes pro-inflammatory prostaglandins, leukotrienes and their primary metabolites, specialized pro-resolving mediators (SPM) including resolvins, protectins, maresins, and lipoxins, biomarkers of ROS and NOS, and fatty acid and mono-hydroxy fatty acid precursors. The positive mode panel includes cysteinyl leukotrienes, PAF, and maresin and protectin conjugates in tissue regeneration (MCTRs and PCTRs). Using standards, MRM transitions were developed and optimized for 88 different lipid mediators, including 16 deuterium labeled internal standards. Next a number of columns were evaluated. The Kinetex Polar C18 column was selected because it provided both polar and non-polar retention that facilitates the separation of lipid mediator epimers and isobaric compounds which would improve assay performance statistics and minimize false positives. As an example, baseline separation of LXA4 and 15-epi-LXA4 epimers, as well as RvD1 and 17-epi-RvD1 epimers was achieved which enables the differentiation of enzymatic pathway utilization for pro-resolving lipid mediator production. Concentration curves were generated by injecting 3 replicate injections of 8 different concentrations of each lipid mediator standard. Excellent linearity is observed with R2 of at least 0.997 for all analytes. Excellent sensitivity was obtained for all lipid mediator species with LLOQ between 0.05 and 1 ng/mL and CV < 30%. Finally, the unique qualitative / quantitative QTRAP platform also enabled MRM triggered EPI experiments which provided high sensitivity MS/MS data for confirmation of low level analytes. The polar C18 LC-MS/MS method provides a unique qualitative / quantitative platform for rapid and sensitive profiling of lipid mediators.

"An Accelerated Data Workflow for High Throughput Serum Lipidomic Studies Using Multisegment Injection-Nonaqueous Capillary Electrophoresis-Mass Spectrometry" (154)


Ritchie Ly, McMaster University (Primary Presenter)

Nicholas Ly, McMaster University

Kazunori Sasaki, Human Metabolome Technologies

Makoto Suzuki, Human Metabolome Technologies

Kenjiro Kami, Human Metabolome Technologies

Yoshiaki Ohashi, Human Metabolome Technologies

Philip Britz-McKibbin, McMaster University


Comprehensive profiling of lipids currently remains a major technical challenge due to their complex chemical structures, variable physicochemical properties and wide dynamic range in biological samples. This issue is further exacerbated when exploring nontargeted approaches for the discovery of clinically significant lipids from thousands of detectable molecular features given the lack of chemical standards and incomplete mass spectral databases for structural elucidation. Current lipidomic studies typically rely on LC-MS methods using reversed-phase and/or hydrophilic interaction-based separations that are constrained by low sample throughput, complicated data workflows and/or limited peak capacity for resolution of diverse lipid classes. Herein, we introduce an accelerated data workflow to resolve a wide range of ionic lipids from serum extracts using multisegment injection-nonaqueous capillary electrophoresis-mass spectrometry (MSI-NACE-MS) under negative ion mode conditions. An iterative two-stage process was developed to rigorously filter out a large fraction of spurious signals, redundant ions and background contaminants from authentic lipids measured from serum extracts when using MSI-NACE-MS with a seven-sample serial injection format in conjunction with a dilution trend filter. This temporal signal pattern recognition strategy allowed for unambiguous authentication of over 200 primarily acidic lipids annotated based on their characteristic accurate mass and electrophoretic mobility. For the first time, we demonstrate that a mobility plot can serve as an effective tool for classifying different serum lipid classes that complements high resolution MS/MS for unknown identification, including non-esterified fatty acids, bile acids and various anionic phospholipids (e.g., phosphatidylserines, phosphatidylinositols, phosphatidic acids etc.). Also, our method was applied to uncover distinctive serum lipid profiles as promising biomarkers associated with liver fibrosis scores from a cohort of patients at advanced stages of non-alcoholic fatty liver disease. In summary, multiplexed separations by MSI-NACE-MS offers a high throughput platform for rapid profiling of polar/ionic lipids with unique selectivity to conventional chromatographic methods, which is optimal for large-scale clinical lipidomic studies with stringent quality control.

"CRAFT for NMR Lipidomics: Targeting Lipid Metabolism in Leucine Supplemented Tumor Bearing Mice" (178)


Hayden Johnson, University of Memphis (Primary Presenter)

Melissa Puppa, University of Memphis

Marie van der Merwe, University of Memphis

Aaryani Tipirneni-Sajja, University of Memphis


Lipid quantification in biological tissues is a powerful tool as lipid dysregulation accompanies a vast number of increasingly common physiological conditions such as diabetes, metabolic syndrome, and cancer. Lipid profiling by 1H-NMR has gained increasing utility in many fields due to its intrinsically quantitative, non-destructive nature and the ability to differentiate small molecules based on their spectral location. Most NMR techniques for metabolite quantification use frequency-domain analysis that involves many user-dependent steps such as phase and baseline correction, and quantification by either manual integration or peak fitting. Recently, Bayesian analysis of time-domain NMR data has been shown to reduce operator bias and increase automation in NMR spectroscopy. These Bayesian methods also been shown to be more effective when utilizing univariate and multivariate statistical analysis – making them particularly suited for the lipidomics context. In this study, we demonstrate the use of CRAFT (Complete Reduction to Amplitude Frequency Table), a Bayesian-based approach to automate processing in NMR-based lipidomics using lipid standards and tissue samples of healthy and tumor bearing mice supplemented with leucine. Complex mixtures of lipid standards were prepared and examined using CRAFT to validate it against conventional Fourier Transform (FT)-NMR and derive fingerprint to be used for analyzing lipid profiles of serum and liver samples. CRAFT and FT-NMR were comparable in accuracy, with CRAFT achieving higher correlation in quantifying several lipid species. The tabular result of CRAFT facilitated fast lipid quantification for all samples. Analysis of serum lipidome of tumor-bearing mice revealed higher phosphatidylcholine, docosahexaenoic acid, polyunsaturated fatty acid, and triglyceride concentrations compared to healthy mice. This serum hyperlipidemia, coupled with no signs of hepatic triglyceride accumulation compared to healthy mice, supported the hypothesis that the tumor-bearing mice were in a state of pre-cachexia at the time of sacrifice. Leucine-supplementation was associated with minimal changes in the lipid profile in both tissues. In conclusion, our study demonstrates that the CRAFT method can accurately identify and quantify lipids in complex lipid mixtures and murine tissue samples. This time-domain approach allowed several bias-inducing steps to be circumvented, subsequently increasing automation and reproducibility in the NMR-based lipidomics workflow.

Tuesday, September 15, 5:00-5:45 EDT

Model Organisms

"UHPLC-MS-Based Lipidomic and Metabonomic Investigation of theMetabolic Phenotypes of Wild Type and Hepatic CYP Reductase Null (HRN) Mice" (25)


Nicola Gray, Murdoch University

Lee Gethings, Waters Corporation

Robert Stephen Plumb, Waters Corporation, Milford, MA, USA (Primary Presenter)

Ian Wilson, Imperial College London


Cytochrome P450 reductase (EC also known as POR (or CPR) is an NADPH-cytochrome P450 oxidoreductase and, as the key electron donor is essential to the functioning of this important superfamily of drug and xenobiotic metabolising enzymes. In the “Hepatic Reductase Null (HRNTM)” mouse the “Cre-lox system” as been employed to selectively delete this enzyme from the liver, but not to affect other tissues. As such, this mouse provides a useful model which can be used to establish the role of the hepatic CYP450 system in xenobiotic metabolism and toxicity studies. Hence, HRNTM mice are deficient in hepatic P450 activities, whereas the activities of other hepatic drug-metabolising enzymes (including UDP-glucuronosyltransferaces; UGTs etc.,) and extrahepatic P450 enzymes are unaffected. HRN mice thus provide a useful in vivo model to define whether or not the majority of CYP450 xenobiotic metabolism is undertaken in the liver, in extrahepatic tissues or a mixture of both. However, the lack of hepatic CYP450s is not without consequences and HRNTM mice show abnormal liver histopathology in comparison to normal mice as well as elevated plasma ALT, GLDH and ALP activities. In order to gain insights into the endogenous metabolic consequences that result from this lack of functional hepatic CYP450s we have compared the tissue metabolic phenotype obtained for the livers of the HRNTM mouse with those of normal control (C57BL6) mice. These metabolic phenotypes were acquired using untargeted UPLC-MS using a two-step extraction method to obtain both polar and non-polar metabolites from extracts of HRNtm mouse liver tissue. Organic tissue extracts were analyzed by RP-UPLC-MS, which utilized a data independent (DIA) strategy. In this mode of operation, function one utilized a low collision energy (4 eV) whilst the second (high energy) function consisted of a collision energy ramp (15-45 eV). Progenesis QI was used for subsequent data analysis to align and normalize the data. Statistical analysis was conducted on the entire data set using EZinfo (Umetrics, Umeå, Sweden) for multivariate statistics, using orthogonal partial least squares-discriminate analysis (OPLS-DA) to establish group differences. Identifications for the differencial markers was achieved by searching against the LipidMaps database.

The results obtained from the UPLC-MS metabonomic and lipidomic profiling of livers from WT and HRNTM mice presented here provide further confirmation of the wide-ranging disruption of normal lipid and bile acid metabolism in POR-deficient animals. Numerous CYP encoded enzymes are involved in sterol metabolism (including cholesterol and bile acids), fatty acids and eicosanoids, evidenced here in the relative quantities of bile acids and disruption in lipid homeostasis, predominantly affecting phosphatidylcholines (PC) and phosphatidylethanolamines (PE).

Previous studies have noted abnormal plasma ALT, GLDH and ALP activities observed in vehicle-treated HRNTM mice, indicative of liver damage or inflammation. These underlying abnormalities in the HRNTM animals are reflected in the phenotypic alterations detected by untargeted LC-MS profiling of liver tissue from WT and HRN™ animals. The integration of metabolomic and lipidomic profiling of liver tissue using a two-phase extraction provides a comprehensive overview of the perturbations in metabolism of the HRNTM mouse.

"Bridging Phenotypical and Analytical Data – Worm Development using Biosorter, Microscopy, and MALDI-MS of C. elegans Cuticle Extracts" (39)


Brianna Morgan Garcia, Department of Chemistry, University of Georgia (Primary Presenter)

Amanda Shaver, University of Georgia

Arthur S Edison, University of Georgia

I. Jonathan Amster, Department of Chemistry, University of Georgia

Franklin E. Leach III, Department of Environmental Health Science, University of Georgia


C. elegans is a free-living nematode and widely used model organism to study development, aging, neurobiology, and behavior. The life cycle of C. elegans is comprised of the embryonic stage, and four larval stages (L1-L4) before reaching adulthood. Synchronized populations of larval stages have been shown to be a powerful tool for RNAi screens, microarrays, sequencing experiments, among others. However, conventional synchronization techniques such as manual picking, gravity stratification and chemical bleaching diminish the ability to understand the animal’s innate physiology, naturally occurring pheromones, and population and chemical dynamics. Herein, we aim to create and validate a method to chemically identity C. elegans life stages by combining large particle flow cytometer phenotypic data of a large mixed-stage population with MALDI-MS metabolite data.

At each larval stage the extracellular cuticle of C. elegans is shed and an entirely new cuticle generated; differing in molecular expression, thickness and composition. 12T FT-ICR MALDI-MS can rapidly generate ultra-high resolution biochemical data from small sample volumes in an automated fashion. Preliminary experiments have shown strong, high signal-to-noise (S/N) MALDI-MS cuticle extract features for as low as 10 worm equivalents in both positive and negative mode. These efforts have also included a variety MALDI matrix compounds. Spectra acquired from 24 averaged scans of 75 laser shots each, with a laser frequency of 200 Hz, resulted in over 200 deconvoluted features. Level 5 (Schymanski scale) accurate mass annotations using MetaboScape and Lipid Maps include molecules such as saturated and unsaturated fatty acids, phosphatidylcholines (PCs), triglycerides (TGs), among others. Combining large particle flow cytometry worm length, optical density measurements, and microscopy imaging with MALDI-MS measurements of cuticle and epicuticle extracts, we can identify stage-specific metabolites and further understand stage-specific changes related to cuticle molting while eliminating the need for synchronization.

"Metabolic profiling of Saccharomyces cerevisiae in response to deletions of genes involved in the glucose repression pathway" (89)


April Miguez, Georgia Institute of Technology (Primary Presenter)

Mark Styczynski


Saccharomyces cerevisiae is a well-established model species of yeast that is of great interest due to its unique metabolism. Most organisms convert glucose into large amounts energy-carrying molecules (ATP) via respiration under aerobic, high glucose conditions, while S. cerevisiae breaks glucose down into ethanol and two ATPs via fermentation. Although we know that the onset of this respirofermentative metabolism is largely controlled by genes involved in the glucose repression pathway, how these genes affect metabolism to initiate this response is still unclear, limiting our ability to effectively and efficiently engineer S. cerevisiae. Here we use a two-dimensional gas chromatography – time of flight mass spectrometer (GCxGC-TOFMS) to investigate the metabolic response of S. cerevisiae to deletions of six key genes involved in this pathway (snf1, reg1, mig1, hxk2, cat8, and hap4). We first determined the growth rates of the deletion mutants and found that the Δreg1 and Δsnf1 strains grow significantly slower than all other strains. We next sought to capture the metabolic effects of the single gene deletions by performing metabolomics analysis on the six mutant strains of S. cerevisiae at log and plateau phase. This study indicated distinct metabolic changes not only due to the temporal variation, but also the genetic variation. We were able to identify unique metabolite changes over time due to each gene deletion mutant; specifically, we found that the Δreg1 and Δsnf1 strains had distinct metabolite profiles from each other. For example, in the Δreg1 mutant, inosine accumulated while guanosine was depleted over time, whereas in the Δsnf1 strain, inosine was depleted over time. These strains are particularly interesting because reg1 and snf1 genes play key, but opposite, roles in the glucose repression pathway. Reg1 represses Snf1 from initiating the expression of respiration genes when glucose levels are high. Extracellular glucose and ethanol were also measured throughout the time course to determine how fermentation and respiration are affected by each knockout, finding that the Δreg1 and the Δhxk2 mutants produce the least amount of ethanol in comparison to the other strains. In future experiments, we will explore ways to validate some our metabolic findings through genetic complementation tests, as well as metabolically engineering the cells to alter ethanol production and to improve overall growth using our metabolomics data to inform our efforts.

Metabolite Identification

"Enzymatic production of xenobiotic metabolites on a high-throughput platform for metabolomic feature identification" (27)


Ken Liu, Clinical Biomarkers Laboratory, Emory University (Primary Presenter)

Grant Singer, Emory University

Choon Lee, Emory University School of Medicine

Edward Thomas Morgan, Emory University

Dean Jones, Emory University School of Medicine


Tens of thousands of features are routinely detected by high resolution mass spectrometry (HRMS) that cannot be identified by conventional MS-based criteria due to their low abundance and the lack of authentic standards. A large fraction of these unknowns is thought to be xenobiotics and their metabolites. To address this problem, we developed biology-based criteria to increase the confidence of metabolite ID, which use metabolites produced in the laboratory from known enzyme systems to match with features of the same retention times and accurate mass on the same HRMS platform. Authentic Phase 1 and Phase 2 metabolites were produced from a range of xenobiotics in a 96-well format using human liver S9 fractions, providing a scalable tool to generate metabolites from thousands of xenobiotics. HRMS detected known and predicted metabolites (using Biotransformer) on the high throughput platform. Selected metabolites were used to support their identification in mouse and human samples based on matched accurate mass MS1, retention time, and co-occurrence with the parent xenobiotic and/or related metabolites in the samples. Generation of metabolites from unlabeled and isotopically labeled precursors also facilitated the discovery of previously unreported biotransformation products. This work was supported by grant U2C ES030163 form the National Institutes of Health.

"Development of an Integrated UHPLC-QTOF-MS/MS-SPE and MicroED Platform for Confident Metabolite Identifications in Metabolomics" (37)


Rajarshi Ghosh, University of Missouri (Primary Presenter)

Guanhong Bu, Arizona State University

Xiaoqing He, University of Missouri

Tommi White, University of Missouri

Brent Nannenga, Arizona State University

Lloyd W. Sumner, University of Missouri


Integration of sophisticated analytical tools into cutting-edge metabolomics platforms is crucial to achieve higher throughput and unambiguous metabolite identifications. In this study, a previously developed UHPLC-QTOF-MS/MS-SPE metabolomics platform was integrated with microcrystal electron diffraction (MicroED), a cryo-EM method capable of higher throughput stereochemical structure elucidation of small molecules. Initially, flavonoid standards such as quercetin and rutin were analyzed by MicroED to evaluate the method’s suitability to confidently determine natural product structures. The structures of both quercetin and rutin were precisely determined by MicroED. Encouraged by our initial MicroED results, we decided to couple the method to UHPLC-QTOF-MS/MS-SPE ensemble established for automated purification and concentration of target analytes. Targeted plant metabolites were purified by splitting a small fraction of the UHPLC eluent towards a MS detector (5%) and the majority fraction to a solid phase extraction (95%) cartridge system. Extracts from Medicago truncatula roots and Fagopyrum esculentum (buckwheat) seeds (100 mg lyophilized plant material extracted with 10 mL 80% methanol) were repeatedly injected (20 times; injection volume = 5 µL) to purify microgram quantities of 6-malonyl ononin and rutin respectively. The purified metabolites were eluted from SPE cartridges using 80% methanol and slowly evaporated at 4 °C for 2 weeks to form microcrystals. The respective MS/MS spectra of the targeted metabolites were compared to that of reference spectra in mass spectral databases for putative identification. Putatively identified metabolites were used for further MicroED experiments. The SPE-purified metabolites were initially screened for the presence of microcrystals using a FEI Tecnai F30 Twin TEM (Transmission Electron Microscope). Continuous rotation electron diffraction data of the SPE-purified microcrystals were then collected from several well-diffracting microcrystals using a Thermo Fisher Scientific Titan Krios cryo-TEM. High resolution diffraction patterns (< 1 Å) were observed from microcrystals of both SPE-purified 6-malonyl ononin and rutin. The diffraction patterns were processed for precise structure elucidation using standard crystallography software suites such as XDS and SHELXT. The proposed UHPLC-QTOF-MS/MS-SPE and MicroED platform offers an exciting and novel approach towards solving the grand challenge of higher-throughput confident metabolite identification in metabolomics.

"Untargeted LC-MS/MS paired with post-acquisition data processing for cyanotoxin detection" (55)


Kim McDonald, Carleton University

Ottawa, Canada


Cyanobacterial blooms have emerged as a major threat to freshwater ecosystems around the world. Eutrophication and climate change are recognized as major drivers of bloom formation. Despite this, the impact most cyanotoxins have on human and ecosystem health are largely unknown. A major challenge for studying cyanotoxins is their immense structural diversity. Currently, cyanotoxin monitoring and management approaches are limited by a lack of reference materials with well-defined toxicological profiles. To study the diversity of cyanotoxins from in vitro cultures and surface water, we have developed data processing approaches to mine untargeted LC-MS/MS data. This enables the categorization of cyanotoxin groups based on specific structural features including both known and new compounds. New putative chemicals with unknown toxicity can be prioritized from in vitro cultures for purification, structural characterization and bioassay studies. We applied our data processing approaches to 5 Microcystis cultures and surface water samples from Ontario and Quebec. Each culture produced different mixtures of microcystins, cyanopeptolins, aeruginosins, cyanobactins, microginins, anabaenopeptins and microviridins. Microcystins and cyanopeptolins were detected with the most frequency from surface water samples. Aeruginosins, microviridins and anabaenopeptins were detected with less frequency. Our research has demonstrated the utility of post-acquisition data processing as a rapid tool to identify different cyanotoxin groups from cyanobacteria cultures and surface water. It also highlights the need to perform more detailed studies on the genetics and toxin production of different cyanobacteria populations across the country.

"Ovarian Cancer Metabolomics: Targeted Microchip Capillary Electrophoresis-Mass Spectrometry to Track Disease Progression." (61)


Samyukta Sah, Georgia Institute of Technology (Primary Presenter)

David Gaul, School of Chemistry and Biochemistry, Georgia Institute of Technology

Eun Young Park, Department of Biochemistry and Molecular Biology, Indiana University School of Medicine, Indianapolis, IN 46202, USA

Olga Kim, Department of Biochemistry and Molecular Biology, Indiana University School of Medicine, Indianapolis, IN 46202, USA

Jaeyeon Kim, Department of Biochemistry and Molecular Biology, Indiana University School of Medicine, and Indiana University Melvin & Bren Simon Cancer Center Indianapolis, IN 46202, USA

Facundo M. Fernández, School of Chemistry and Biochemistry, Georgia Institute of Technology


Ovarian cancer (OC) is the most lethal gynecologic malignancy and is the fifth leading cause of death by cancer among women in the world. High-grade serous ovarian cancer (HGSC), an OC subtype accounts for 70-80% of the OC deaths, and in 80% of the cases is diagnosed in the advanced stages. The difficulty in detecting HGSC at early stages, and consequently the high mortality rates, mainly arises from non-specific symptoms. Mass spectrometry (MS)-based metabolomics has been essential in discovering key metabolites associated with the disease, leading to a better understanding of pathophysiology and enhanced diagnostics.

Here, microchip capillary electrophoresis (µCE, 908 Devices) coupled with Orbitrap MS (Thermo) was used to develop an assay for monitoring HGSC progression and identifying metabolites indicative of early stage HGSC. This µCE-MS method reduced the analysis time to less than 3 minutes, much faster than traditional chromatography, and required only 4 nL of sample per injection, enabling sampling serum from animals more often. To better understand the pathogenesis of OC, we developed two mouse models of HGSC: (1) a double-knockout (DKO) mice (Dicer1 flox/flox Pten flox/flox Amhr2 cre/+) by inactivating the Dicer1 and Pten genes, and (2) a triple-mutant (TKO) mice (p53 LSL-R172H/+ Dicer1 flox/flox Pten flox/flox Amhr2 cre/+) by adding a p53 mutation (R172H), which is equivalent to human p53-R175H mutant, one of the frequent p53 mutations found in human OC.

A 2-minute assay was developed for 12 metabolites in a scheduled and multiplexed parallel reaction monitoring (PRM) mode. Peak extraction was done using Skyline (v19.1), and the peak area %RSD (n =4) ranged from 10-29% and migration time %RSD was 3-5%. Detection limits for metabolites such as methionine, normetanephrine, and arginine were 3nM, 5nM and 10nM respectively. However, extending the method to more than 12 metabolites showed that in some cases, the Orbitrap mass analyzer was not fast enough to put sufficient points across the analyte peaks due to their short duration (1-2 seconds). Therefore, a full scan method for 41 metabolites was developed. The separation was complete within 2-minutes; some metabolites coeluted but they could be distinguished by their m/z values in Skyline (v19.1). Additionally, structural isomers such as leucine and isoleucine were baseline resolved. For this full scan method, calibration curves (OriginPro 2020) were plotted using a weighted linear regression method (1/x2). The R2 values ranged between 0.94-0.99 for the concentration range analyzed, and the detection limit ranged from (3-155nM). Metabolites with quaternary amine groups such as carnitine and betaine were observed to have low detection limits of 3nM and 4nM respectively. The instrument stability was also monitored for all measurements by adding 13C-phenylalanine as an internal standard. The %RSD (n = 20) for 13C-phenylalanine was 22%. The method performance was further evaluated by spiking mice serum samples with known amount of standard. The method was used to analyze 30 metabolites in serum samples of the triple-mutant (TKO) mouse model. Next, longitudinal samples will be analyzed to monitor disease progression to identify potential early stage HGSC biomarkers.

"Alzheimer’s disease and mild cognitive impairment" (65)


Ali Yilmaz, Beaumont Research Institute; Oakland University-William Beaumont School of Medicine (Primary Presenter)


The lack of sensitive and specific biomarkers for the early detection of Mild cognitive impairment (MCI) and Alzheimer’s disease (AD) is a major hurdle to improve patient management. A targeted, quantitative metabolomics approach using both 1H NMR and mass spectrometry were employed to investigate the performance of urine metabolites as potential biomarkers for MCI and AD. Correlation-based feature selection (CFS) and Least absolute shrinkage and selection operator (LASSO) were used to develop biomarker panels tested using support vector machine (SVM) and logistic regression models for diagnosing each disease state. Metabolic changes were investigated to identify which biochemical pathways were perturbed as a direct result of MCI and AD in urine. Using SVM, we developed a model with 94% sensitivity, 78% specificity and 78% AUC for distinguishing healthy controls from AD sufferers. Using logistic regression, we developed a model with 85% sensitivity, specificity of 86% and an AUC of 82% for AD diagnosis as compared to cognitively healthy controls. Further, we identified 11 urinary metabolites significantly altered to include: glucose, guanidinoacetate, urocanate, hippuric acid, cytosine, 2- and 3-hydroxyisovalerate, 2-ketoisovalerate, tryptophan, and malonate in AD patients which are also capable of diagnosing MCI, with a sensitivity of 76%, specificity of 75%, and accuracy of 81% as compared to healthy controls. This pilot study suggests that urine metabolomics may be useful for developing a test capable of diagnosing MCI and AD from cognitively healthy controls.

Comparing accurate mass MS/MS spectral similarity algorithms for small molecules" (71)


Yuanyue Li, UC Davis (Primary Presenter)

Tobias Kind, UC Davis Genome Center - Metabolomics

Oliver Fiehn, UC Davis


MS/MS spectral similarity comparison is a fundamental method in mass spectrometry data analysis, yet surprisingly little is known how well different algorithms perform. In untargeted metabolomics and lipidomics, compound identification strategies strongly rely on similarity match scores between experimentally acquired spectra and experimental or in silico library spectra. Experimental MS/MS spectra may differ from library spectra by use of different mass spectrometers, different fragmentation energies and different data processing parameters such as signal-to-noise ratios (S/N), isotope deletion and abundance cut-offs. Hence, different experimental spectra, and perhaps, different compound classes, may favor different similarity comparison algorithms. To contrast performance of MS/MS matching approaches, use of comprehensive MS/MS libraries is needed for benchmarking.

We reviewed more than 60 distance metrics and similarity algorithms and implemented 39 methods in Python, along with the NumPy library, for use in benchmarking tests. Algorithms extended classic dot-product similarities and methods used in MS/MS identification software such as MSforID and NIST MS software to include a range of further matrix approaches. These algorithms included vector-based, probability-based and entropy information theory-based similarity. Performance was tested on NIST17.

We have successfully tested 39 different algorithms, grouped by assumptions and axioms underlying the different methods. For example, vector-based algorithms transfer MS/MS spectral data into intensity vectors and then compare distances between pairs of intensity vectors. In contrast, probability-based algorithms suppose that the intensity of a fragment ion represents the probability of the detection of a specific product ion and subsequently uses the overlap of pairs of MS/MS probability distributions as similarity score. A third group of algorithms uses entropy information theory. Such similarity scores use entropy to represent the overlap of information revealed by MS/MS spectra comparisons. Classic algorithms consider simpler ways of MS/MS matching. The MSforID algorithm compares the number of matched ions between two spectra, matching ion m/z and intensity values. The NIST algorithm is based on dot-product similarities, using mass weight-scaled intensities that give additional weight to relative peak intensities for spectra that share many fragment ions. Parameters in both the MSforID and NIST algorithms were optimized with multiple experiment data to maximize identification performances.

Overall, more than 200,000 mass spectra of small molecules were tested against each of the MS/MS similarity scoring algorithms. Results will be shown concerning the accuracy, sensitivity, and selectivity for these methods, separated by instrument classes (QTOFs and Orbital Ion Traps) and metabolite compound classes.

"Is High-Resolution Fourier Transform Ion Cyclotron Resonance Mass Spectrometry Needed to Improve Metabolite Annotation?" (77)


Danning Huang, Georgia Institute of Technology (Primary Presenter)

Marcos Bouza Areces, Georgia Tech

David Gaul, School of Chemistry and Biochemistry, Georgia Institute of Technology

Arthur S Edison, University of Georgia

Facundo M. Fernández, School of Chemistry and Biochemistry, Georgia Institute of Technology


Fourier transform ion cyclotron resonance (FT-ICR) and Orbitrap mass spectrometry (MS) are the highest-performing analytical platforms in metabolomics. Their high mass accuracy and resolution enables detailed investigation of biological metabolomes. Non-targeted MS experiments, however, yield extremely complex datasets that make metabolite identification very challenging. High resolution accurate mass measurements greatly facilitate this process by reducing mass errors and spectral overlaps. When applied together with isotope pattern matches, heuristic rules, and restrictions during searches, the number of candidate empirical formula(s) found can be significantly reduced. Here, we evaluate the quality of the mass measurements produced by FT-ICR and high field Orbitrap platforms, and how these affect the assignment of the correct elemental formulae in metabolite identification applications.

In this study, a pooled metabolite sample was prepared by mixing 104 metabolite standards with a final concentration of 5μM in methanol. The pooled metabolite samples were subjected to direct infusion MS analysis on a Bruker 12-T solariX FT-ICR mass spectrometer, and HILIC UPLC-MS analysis on a Thermofisher Scientific Orbitrap ID-X Tribrid mass spectrometer, in both positive and negative ion modes. Mass accuracy and elemental formulae assignment were evaluated using different combinations of automatic gain control (AGC: 5e4, 1e5, and 5e5) and resolution settings (R: 120K, 240K, and 500K) on the Orbitrap ID-X mass spectrometer. Experimental conditions for both platforms were optimized for coverage of the standard metabolites detected.

Following data acquisition, MZmine 2.51 was used for processing of sample data generated from both platforms. Empirical formula prediction of the extracted metabolite features was performed with a < 1 ppm (or 0.001 Da) mass error tolerance. [C0-50H0-100N0-15O0-20P0-7S0-8±H] ± limits, heuristic rules, and isotope pattern matches were also applied through built-in functions. In positive ion mode, 75 metabolites were detected by FT-ICR mass spectrometer (size: 8M, R: ~500K), and 70 metabolites were detected by Orbitrap ID-X mass spectrometer using equivalent resolution settings (R: 500K, AGC: 1e5). The average mass error for the correctly assigned metabolites was 0.099 ppm and 0.369 ppm on the FT-ICR and Orbitrap ID-X mass spectrometers, respectively. Effectiveness of correct elemental formulae assignment was evaluated. By only using mass accuracy as the criterium, 100% (75/75) and 90% (63/70) of metabolites were assigned the correct formulae as the rank 1 candidate using the FT-ICR and Orbitrap ID-X mass spectrometers, respectively. After adding the isotope pattern score additionally to the mass accuracy, the performance of correct formula rank #1 assignments improved to 94.3% (66/70) for the Orbitrap ID-X. The same data analysis workflow was also applied to datasets collected on the Orbitrap ID-X mass spectrometer using different combinations of AGC and R settings as well as datasets collected in negative ion mode. This study provides the first comparison between high-resolution FT-ICR-MS and Orbitrap Tribrid MS platforms for elemental formula annotation in metabolomics.

"Correlation analysis of a complete dataset to determine fragments and adducts" (129)


Chris Beecher, IROA Technologies (Primary Presenter)

Alexander Raskind, University of Michigan Compound Identification Development Core

Felice de Jong, IROA Technologies


In order to determine if the majority of unknowns in Metabolomics are really fragments or adducts we have added a feature that creates examines the correlational structure between all of the compounds found in an IROA-based Metabolomics study. The use of IROA at the start assures that all of the compounds selected for analysis represent real biological data, These IROA peaks identified across all samples are binned by common features and then the features are examined for their correlation as a function of time. We find that most highly correlated structures represent chemically understandable relationships. This software is built into the ClusterFinder software beginning in version 4. The single function examines an entire multi-sample dataset and uses all of the information from all samples to find and score correlations. The use of more samples in the analysis increases the resolution of the relationships that can be found. We will report preliminary statistics in this presentation.

"Metabolic changes in murine hair follicles treated with Procyanidine-B2 rich nutraceuticals studied by Magnetic Resonance Mass Spectrometry (MRMS)" (139)


Eduardo Sommella, University of Salerno, Fisciano, Italy

Emanuela Salviati, University of Salerno, Fisciano, Italy

Matthias Witt, Bruker Daltonik GmbH, Bremen, Germany

Christopher J Thompson, Bruker Daltonics (Primary Presenter)

Pietro Campiglia, University of Salerno, Fisciano, Italy



Known for anti-inflammatory and antioxidant properties, nutraceuticals enriched in Procyanidin-B2 promote hair growth both in vitro and in vivo. However, the metabolic changes associated with the treatment have not been elucidated. In this study, flow injection analysis magnetic resonance mass spectrometry (FIA-MRMS) was employed to understand the metabolic shift produced by treatment with Procyanidin-B2 nutraceuticals (Annurca apple extract) in murine models. FIA-MRMS allowed the identification of several metabolites using ultra-high mass accuracy and fast analysis time. Glutaminolysis, pentose phosphate pathway, glutathione, citrulline and nucleotide synthesis derived metabolite were detected. The metabolic profile revealed that the treatment with Procyanidin-B2 results in the early exit of hair follicles from telogen phase and increased keratin biosynthesis.


Wild-type C57/BL6 mice (7 weeks old, postnatal day 42) were used in all experiments to test the effect of cosmetic foam containing Annurca apple extract (AAE). Mice tissues were rinsed and kept in PBS immediately after tissue excision. Hair shafts were plunked out with sterile tweezers and immediately covered with a solution of PBS at room temperature. The cell pellets were washed twice in PBS and homogenized in 1 ml of pre-chilled methanol/water 80:20 solution and finally centrifuged at 10,000 g for 10 min at 4°C. Analyses were performed by flow injection analysis Electrospray ionization at a flow rate of 2 μL/min. Data were acquired on a MRMS solariX XR 7T system (Bruker Daltonik GmbH, Bremen, Germany).

Preliminary Data

C57/BL6 mice were topically treated with a foam supplemented either with AAE or with a placebo. After 4 weeks of treatment, mice (11 weeks old) were sacrificed and their dorsal skin was excised. Skin biopsies were embedded in paraffin and prepared for histology. The metabolic content of hair follicle cells plucked out by mice were treated topically with AAE and analyzed by DI-MRMS mass spectrometry. By screening intracellular metabolites with similar alteration tendency in all the AAE treated mice, glutaminolysis, pentose phosphate pathway (PPP), amino acid oxidation, mitochondrial β-oxidation as well as arginine metabolism became our focus. Significant elevation of glutamine and glycine as well as the increase in the intracellular level of the PPP intermediate ribulose-5-phoshate together with the reduction of the intracellular level of nucleotides and deoxy-nucleotides suggest that AAE causes a reduction in the utilization of glucose and glutamine for PPP. This is a metabolic pathway that correlates with increased keratin biosynthesis in hair follicles. Additionally, reduced intracellular level of glutathione also confirmed that the catabolism of glutamine is halted in AAE treated hair follicles.

"More Confident Metabolomics Identifications: High-Resolution Ion Mobility Separations with Structures for Lossless Ion Manipulations (SLIM)" (141)


Chris Conant, Pacific Northwest National Laboratory (Primary Presenter)

Kent Bloodsworth, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA

Daniel Orton, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA

Aivett Bilboa, Environmental Molecular Sciences Laboratory, Pacific Northwest National Laboratory, Richland, WA

Joon-Yong Lee, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA

Yehia Ibrahim, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA

Ailin Li, Pacific Northwest National Laboratory

Xueyun Zheng, Pacific Northwest National Laboratory

Richard Smith, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA

Thomas Metz, Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA


The field of metabolomics has experienced significant growth in the last decade due to increased recognition of the biological importance of small molecules, as changes in the metabolome have proven to be effective for detecting and characterizing disruption of upstream biological processes. However, the structural diversity of metabolites makes identification difficult. Metabolomics studies typically rely on mass spectrometry (aided by front-end gas or liquid chromatography separations) for molecular identification, but the fragmentation patterns of these small molecules are often ambiguous.

Ion mobility spectrometry-mass spectrometry (IMS-MS) is finding increasing usage in metabolomics separations. IMS separates ions based on differences in their collision cross sections (CCS) and provides information on ion shape and structure. These separations are orders of magnitude faster (milliseconds to seconds) than liquid chromatography (LC) separations, representing an attractive alternative for increased throughput, but are also ideally situated for multidimensional analysis as a hyphenated LC-IMS-MS technique. The recent development of traveling wave (TW)-based Structures for Lossless Ion Manipulations (SLIM) IMS has pushed the limits for high-resolution ion mobility measurements owing to significant developments in ion confinement and extended path length. TW-IMS is widely used for CCS determination, but unlike classical drift tube (DT)-IMS, which allows for direct CCS calculation based on ion arrival times, TW-IMS relies on internal calibrants having CCS values measured on, e.g., DT-IMS systems. The accuracy of such calibration is often somewhat sensitive to experimental conditions, as well as the structural similarity of calibrants and analytes. Thus, characterizing the effects of TW conditions or calibrant-choice towards minimizing calibration error is crucial for establishing a foundation for TW-IMS CCS measurements and for making measurements useful between different IMS platforms.

Here we compare CCS values measured on a commercial DT-IMS system with CCS measurements made on a SLIM TW-IMS platform for the ~450 compound IROA metabolite set. Solutions containing these metabolites were introduced to either instrument by nanoelectrospray ionization, and separations were carried out in ~3 Torr Helium buffer gas. Aided by our inhouse AutoCCS tool, DT-IMS CCS values were calculated by a step-field method, and TW-IMS CCS were determined from CCS calibrations based on DT-IMS values. Initial experiments found that CCS accuracy was dependent on TW amplitude, speed, and choice of calibrant. Full characterization of the IROA metabolite set was carried out under conditions optimized for accuracy. We found that some metabolites, detected as a single peak in DT-IMS, were resolved into multiple features in the high-resolution SLIM TW-IMS separations, highlighting the utility of such measurements and posing a challenge for the current calibration-based paradigms. Strategies towards independence of TW-IMS from DT-IMS measurements, and for utilizing the high precision feasible with SLIM TW-IMS measurements are presented.

"MetCraft: Pipeline for Metabolomics Data Analysis" (165)


Habtom Ressom, OmicsCraft LLC (Primary Presenter)

Linge Yan, OmicsCraft

Mohammad Ras Nezami Ranjbar, OmicsCraft


Despite a large accumulation of metabolomics data acquired by liquid chromatography-mass spectrometry (LC-MS), effective use of these data in the growing systems biology approaches for metabolite biomarker discovery has been very limited. This is in part due to the lack of effective tools to: (1) accurately determine the identity of many analytes detected by LC-MS; (2) investigate the rewiring and conserved interactions among metabolites in disease vs. control groups; (3) integrate metabolomics data with other omics data to evaluate the relationship between metabolites and disease at the systems level; and (4) perform multiple data analysis steps involved in a typical metabolite biomarker discovery study by LC-MS-based untargeted metabolomics.

We developed MetCraft (, a cloud-based platform with an interactive modular interface, for analysis of data acquired by LC-MS-based metabolomics. Specifically, the platform allows users to build data analytics pipelines through a web-browser and perform data analysis on a remote machine. The pipelines are built by assembling several modules implemented in MetCrfat for identification and ranking of putative metabolite IDs, differential analysis of metabolite profiles, and multi-omics data integration. The modules include the following:

Mass-Based Search Module that searches for putative metabolite IDs in several major compound databases based on m/z values. The module retrieves chemical identifier information for the metabolites by cross-referencing multiple databases.

IF-THEN Decision Making Module that allows users to take advantage of expert-based decision making rules to reduce the number of repetitive and unlikely putative metabolite IDs.

Isotopic Pattern Analysis Module evaluates putative IDs with different elemental formulas by comparing the observed isotopic patterns against the theoretical isotopic patterns.

Network-Based Analysis Module ranks putative IDs by reconstructing metabolic networks and using the resulting network topology for assigning a priority score to each putative ID. The assumption is that a putative ID is more likely if other metabolites that are connected or are in close proximity to it are also detected in the samples from the same study.

MS/MS-Based Search Module allows users to upload MS/MS spectra either in mzXML or plain text format for spectral matching. The module searches for matching MS/MS spectra against libraries that consist of experimental and in-silico fragmentation patterns acquired from various sources including MoNA and NIST.

Differential Analysis Module uses a network-based method to select disease-associated metabolites by combining statistical significance of individual metabolites and differential correlation between metabolite pairs in disease vs. control groups.

Omics Integration Module integrates metabolomics data with other omics datasets to investigate disease-associated metabolites at the systems level. After an intra-omics network is constructed based on metabolomics data, the module calculates differential canonical correlation to evaluate inter-omics interactions.

We believe MetCraft will contribute to systems biology approaches for metabolite biomarker discovery by addressing the major challenges in metabolomics including metabolite identification and multi-omics integration. Furthermore, the pipeline facilitates analysis of untargeted metabolomics data via a cloud-based platform with an interactive modular interface that allows users to build data analytics pipelines by linking modules and run the pipelines remotely.

Microbiome Applications

"Investigating aquacultured sablefish metabolomes using NMR to understand microbiome function and improve larval rearing methods" (91)


Erik Andersson, University of Illinois at Chicago (Primary Presenter)

Melissa Pierce, University of Illinois at Chicago

Jonathan Lee, National Oceanic and Atmospheric Administration

Rachel Poretsky, University of Illinois at Chicago


Sablefish is an economically important fishery in the U.S., but sablefish aquaculture is impeded by high costs and low survival during the larval rearing stage. The addition of an algal supplement (greenwater) can increase larval survival rates, but the additive is expensive and hinders aquaculture efficiency. Substituting greenwater with a cheaper clay-based substitute (claywater) after one week similarly improves survival while decreasing costs. Mechanisms underpinning survival differences are unclear but are important for the development of new or optimized additives to improve aquaculture efficiency of sablefish and other commercially important fish reared using greenwater. Previous results suggest microbial community composition and function may be linked to larval survival rates. For example, we conducted experiments to determine the effect of different additive treatments on microbial composition and communities were taxonomically similar among the water-additive treatments with increased survival. However, the functional contributions of different microbial taxa are not well understood. To clarify the role of microbial communities in larval survival differences we supplement new 16S rRNA gene sequencing results with metabolomics analyses to directly investigate functional differences in larval sablefish microbiomes reared in greenwater (n=6), claywater (n=6), greenwater/claywater mix (n=6) and greenwater exudate (n=6) additives. An untargeted, NMR-based metabolomics approach was used to characterize sablefish metabolic profiles and to identify candidate metabolic pathways that may differ between treatments. A targeted, mass-spectrometry based approach was also used to quantify individual metabolites identified from untargeted analyses or through a priori knowledge. For example, the algal-derived metabolite dimethylsulfoniopropionate (DMSP) and its derivatives were targeted because of previous links to improved larval sablefish survival. The use of these complementary data types provides unique insights into metabolic activities of microbial communities under different water-additive treatments and may produce promising leads linking specific metabolic pathways to survival. These results progress aquaculture practices of the economically important sablefish and help clarify host-environment-microbiome interactions. Here we present the 1H NMR metabolomic profiles of sablefish larvae, in addition to profiles of the rearing tank seawater (exometabolites) and biofilms which help to clarify the metabolic environments experienced by the larvae.

"Fusobacterium spp impart global metabolic dysregulation in Oral cell carcinoma" (153)


Iqbal Mahmud, University of Florida (Primary Presenter)

Timothy Garrett, Univ of FL-Pathology


Oral carcinoma accounts for 10,860 estimated deaths and 53,000 new cases in USA in 2019. It comprises a heterogeneous mixture of tumor cells, nontransformed cells, and distinct microbial community. Fusobacterium spp is among the most prevalent bacterial species with increased abundance upon the progression of oral cancer. Several mechanisms including incorporation of bacterial substances as exposomes, sensitizing host inflammatory mediators, activation of cell proliferative machinery, inhibition of cell death signaling, chemo-therapy resistance and disease recurrence have been suggested about the role of Fusobacterium spp in the pathogenesis of multiple cancers. Surprisingly, contributions of this pathogenic bacterium to cancer metabolism are largely unknown. Using untargeted metabolomics, we studied the metabolic landscape among oral cancer cells, Fusobacterium spp and its infected host cells. Our study provides details of the metabolic landscape that Fusobacterium spp modulates in host cancer cells.

Briefly, untargeted metabolome profiling revealed that Fusobacterium spp contributes to a heterogeneous metabolic signature alteration across the oral cancer cells: First, we evaluated the metabolic similarities and uniqueness among oral cancer cell lines or Fusobacterium spp or Fusobacterium spp infected oral cancer cells using a hierarchical clustering approach. We identified three major clusters of metabolic landscape for each case. For example, in the case of oral cancer cells, RPMI cells showed a distinct metabolic signature from the other cell lines. On the other hand, among the Fusobacterium spp, F. vincentii showed a clearly separate cluster from F. nucleatum and F. periodonticum. Notably, Fusobacterium spp infected oral cancer cells were found to compromise with host metabolic networks associated with bioenergetics and redox balance. Strikingly, our study identified that known genetic vulnerabilities of cancer are associated with Fusobacterium invasion and persistence while modulating to host metabolic networks. Importantly, we have seen over 80% of metabolic features as unidentified in Fusobacterium spp infected oral cancer cells and indicates that Fusobacterium spp may incorporate important metabolic substances as exposome, which may involve oral carcinogenesis and needs to be explored further. This study unravels the comprehensive metabolic resources through which Fusobacterium spp contribute to oral carcinogenesis. Together, these findings provide a comprehensive metabolic landscape that will help to develop the management strategies to prevent Fusobacterium spp mediated oral cancer.


"The Biochemical Profile of Post-Mortem Brain from People Who Suffered from Epilepsy Reveals Novel Insights into the Etiopathogenesis of the Disease" (5)


Ashna Lalwani, Department of Computational Medicine and Bioinformatics, University of Michigan (Primary Presenter)

Ali Yilmaz, Beaumont Research Institute; Oakland University-William Beaumont School of Medicine

Sumeyya Akyol, Beaumont Health System-Research Institute

Stewart F. Graham, Beaumont Research Institute; Oakland University-William Beaumont School of Medicine

Zafer Ugur, Beaumont Health System-Research Institute

Halil Bisgin, Department of Computer Science, Engineering, and Physics, University of Michigan


Epilepsy not-otherwise-specified (ENOS) is one of the most common causes of chronic disorders impacting human health, with complex multifactorial etiology and clinical presentation. Understanding the metabolic processes associated with the disorder may aid in the discovery of preventive and therapeutic measures. Post-mortem brain samples were harvested from the frontal cortex (BA8/46) of people diagnosed with ENOS cases (n= 15) and age- and sex-matched control subjects (n = 15). We employed a targeted metabolomics approach using a combination of proton nuclear magnetic resonance (1H-NMR) and direct injection/liquid chromatography tandem mass spectrometry (DI/LC-MS/MS). We accurately identified and quantified 72 metabolites using 1H-NMR and 159 using DI/LC-MS/MS. Among the 212 detected metabolites, 14 showed significant concentration changes between ENOS cases and controls (p < 0.05; q < 0.05). Of these, adenosine monophosphate and O-acetylcholine were the most commonly selected metabolites used to develop predictive models capable of discriminating between ENOS and unaffected controls. Metabolomic set enrichment analysis identified ethanol degradation, butyrate metabolism and the mitochondrial beta-oxidation of fatty acids as the top three significantly perturbed metabolic pathways. We report, for the first time, the metabolomic profiling of postmortem brain tissue form patients who died from epilepsy. These findings can potentially expand upon the complex etiopathogenesis and help identify key predictive biomarkers of ENOS.

"RaMP - Relational Database of Metabolic Pathways: Integrating metabolomics data with other omics to gain supported insights into biological mechanisms and phenotypes" (47)


John Christian Braisted, National Center for Advancing Translational Science, National Institutes of Health (Primary Presenter)

Andrew Christopher Patt, The Ohio State University

Tara Eicher, National Center for Advancing Translational Science

Ewy Mathe, National Center for Advancing Translational Science


Omics data are increasingly collected in biomedical research to identify mechanisms and biomarkers of disease progression and treatment response. While the utility of omics integration for uncovering novel dysregulated biological and mechanistic pathways has been shown in various contexts, the analysis and interpretation of these data remains complex. Metabolomics provides insights into cellular function and a variety of points of regulation (e.g. transcription and alternative splicing, translational and post-translational modifications, allosteric modulation of enzyme function, etc.) that are impacted by experimental treatments. Metabolomics thus represent functional endpoints of upstream changes in gene expression and protein levels. The combination of metabolite, gene, protein, and other molecular entities (e.g. microbiome) provides a holistic view of a biological system. The power of combining data from multiple omics levels is that findings may corroborate to provide insights that support both regulatory mechanisms as well as functional outcomes.

We present here on RaMP (Relational database of Metabolic Pathways), a platform for interrogating biological roles and pathways based on metabolomics coupled with transcriptomics or proteomics data sets. RaMP consists of a central database of biological pathway and functional annotations associated with genes and metabolites. Managed data resources, frequent updates and extensibility of the knowledge base ensures that RaMP will stay relevant as new annotations and data sources become available. RaMP currently encompasses pathway and role annotations from KEGG, Reactome, WikiPathways, and HMDB. RaMP supports single gene or metabolite queries and batch queries to report on enriched biological themes. Providing metabolite and gene lists together allows simultaneous analysis resulting in reported biological roles and pathways supported by both input types.

We will present on RaMPs infrastructure, current and proposed future data sources, support for running local installations as well as using our public cloud instance. Example input and results will illustrate the tool’s typical workflow and demonstrate utility in uniting omics and metabolomics analysis. Our hope is to elicit feedback and ideas to help improve the tool and advance the interpretation of metabolomic data sets.

"Untargeted Metabolomic Profiling Identifies the Biochemical Signature of Urocanase Deficiency in Amniotic Fluid" (63)


Jing Xiao, Baylor College of Medicine (Primary Presenter)

Adam D. Kennedy, Metabolon, Inc.

Qin Sun, Baylor College of Medicine

V. Reid Sutton, Baylor College of Medicine

Sarah Elsea, Baylor College of Medicine


Background: Untargeted metabolomic profiling provides a powerful tool to study the pathophysiology of metabolic disorders in a function level. While it has been successfully applied to detect the substrates, intermediates, and products of numerous metabolic disorders in plasma, urine, and cerebrospinal fluid specimens, the application of metabolomic profiling in amniotic fluid remains at “embryonic” stage. In this study, we sought to demonstrate that untargeted metabolomic profiling can also identify the phenotypic fingerprinting of a metabolic disorder, urocanase deficiency, in amniotic fluid samples.

Methods: A population-based analysis was performed including thirty randomized amniotic fluid samples which underwent untargeted metabolomic analyses using ultra high-performance liquid chromatography tandemly linked to accurate mass spectrometers.

Results: Pathway analysis of the metabolomic profiles of the thirty amniotic fluid samples identified the biochemical signature associated with urocanase deficiency in one patient sample. The elevations of cis-urocanate (Z-score +4.8), trans-urocanate (Z-score +3.5), and imidazole propionate (Z-score +2.9) are consistent with deficient UROC1 activity.

Conclusions: We describe the unique biochemical signature of urocanase deficiency previously identified in plasma and urine samples from subjects diagnosed with urocanase deficiency. Our results suggest that untargeted metabolomic profiling of amniotic fluid has the potential capacity to identify molecular signatures associated with metabolic diseases.

"Integrated analysis of lipidomics and proteomics data reveals biomolecular processes associated with bladder cancer" (79)


Nobal Dhruw, Elucidata

Abhishek Jha, Elucidata (Primary Presenter)

Lee Gethings, Waters Corporation

David Heywood, Waters Corporation

Andrew Peck, Waters Corporation

Suraj Dhungana, Waters Corporation

Soumya Luthra, Elucidata

Sandhya Srinivasan, Elucidata

Shefali Lathwal, Elucidata



Bladder cancer is one of the most common diseases worldwide, with unknown pathogenesis and a high progression rate. To understand the mechanism of how this progresses, we applied a multi-omics approach to gain a comprehensive understanding of biomolecular processes associated with bladder cancer. Lipidomics, when studied in conjunction with the functional proteomics approach enables the researcher to develop a holistic understanding of the biological question under investigation as this enables a systems biology approach which can help explain the molecular processes that mediate cellular physiology as both lipids and proteins act together to drive these cellular processes through enzymatic activity and signaling.


LipidQuan, a robust platform to carry out rapid targeted quantification of a broad range of lipids, was utilized for the profiling of plasma samples, and Progenesis was used to generate quantitative proteomics data. Downstream statistical analysis, visualization, and integrated-omics studies were carried out on Elucidata’s Polly platform.


Plasma samples were derived out of 12 subjects, of which 6 were healthy controls and the other 6 subjects were patients previously diagnosed with bladder cancer. The quantitative data derived from LipidQuan and Progenesis were processed using Polly workflows. Differential analysis between the tumor and the normal samples show a total of 94 lipid species and 181 proteins as statistically relevant with an adjusted p-value of less than 5%.


Statistical analysis for the lipidomics dataset reveals that ceramides, LPCs, PCs and SMs are differentially expressed in bladder cancer subjects as compared to the control samples; while the proteomics data suggests differential regulation of proteins related to immunoglobulin light chains. Multi-omic approach for this problem unveils the statistically significant dysregulated pathways that consist of glycerolipid metabolism, sphingolipid signaling pathway, and fat absorption and digestion pathway.

"The MultiOmics Explainer: Explaining Omics Results in the Context of a Pathway Database" (101)


Peter Karp, SRI International (Primary Presenter)

Suzanne Paley, SRI International


Which of the results from a high throughput molecular experiment can be explained by existing knowledge about an organism's biological networks, and which results cannot be explained by existing knowledge? Given the large amount of network knowledge now in hand (for example, the EcoCyc database now contains 2800 substrate-level enzyme regulation events, 3600 transcriptional regulatory interactions, and 2000 metabolic reactions), these questions are not easy to answer. The MultiOmics Explainer is a new component of the Pathway Tools software that seeks to answer those questions. The inputs to the tool are a combination of transcriptomics, proteomics, and metabolomics data. The MultiOmics Explainer queries a Pathway/Genome Database such as EcoCyc, searching through the network of metabolic reactions, transport events, cofactors, enzyme substrate-level regulation events, and transcriptional and translational regulation. The software searches for mechanistic connections among its input genes, proteins, and metabolites. Computed connections are displayed in a network diagram that combines all of the preceding types of influences into an explanatory graph. We have tested the software on Escherichia coli experimental data in which changes in metabolite levels resulting from gene knock-outs have been measured. In a number of cases the software is able to generate reasonable explanations of how the gene knock-out would cause the altered metabolite levels.



Sreenath Nair, St. Jude Children's Research Hospital (Primary Presenter)


Human genome wide screening studies have recently identified mutations in SLC29A3 gene that encodes the protein ENT3 (Equilibrative Nucleoside Transporter 3) causes a plethora of autosomal recessive disorders characterized by hyperpigmentation, hypertrichosis, histiocytosis, hepatomegaly and short stature collectively termed as ENT3 spectrum disorders. Although, ENT3 is well known for its nucleoside transport activity with high expression in liver, kidney and placenta; little, if anything, is known about the effects of inactivation of SLC29A3 on the tissue metabolome. In this study, we employed an unbiased mass spectrometry-based metabolomics profiling approach to ascertain the global metabolic changes in SLC29A3-/- liver. We found several physiologically important metabolites belonging to different classes including fatty acid derivatives, vitamins, steroids, carnitines, organic acids, benzenoids, amines and aminoacids, nucleosides and nucleotides altered in SLC29A3-/- liver samples. Although amines, nucleosides and nucleotides have been known as substrates of ENT3 transport, other compounds have not been previously linked to ENT3 transport. Further, targeted lipidomics of liver samples confirmed perturbation in lipid homeostasis in SLC29A3-/- mice, thus, identifying a new role for ENT3 in cellular physiology.


1. Nair, S., Strohecker, A.M., Persaud, A.K. et al. (2019). Nat Commun,10, 2943.

New Experimental Methods & Technologies

"Rapid Biomonitoring of Perfluoroalkyl Substance Exposures in Serum by Multisegment Injection-Nonaqueous Capillary Electrophoresis-Tandem Mass Spectrometry" (23)


Sandi Azab, Department of Chemistry and Chemical Biology, McMaster University (Primary Presenter)

Philip Britz-McKibbin, McMaster University

Rebecca Hum, McMaster University


Perfluoroalkyl substances (PFASs) are a major contaminant class due to their ubiquitous prevalence, persistence and putative endocrine disrupting activity that may contribute to chronic disease risk notably with exposures early in life. Herein, multisegment injection-nonaqueous capillary electrophoresis-tandem mass spectrometry (MSI-NACE-MS/MS) is introduced as a high throughput approach for PFAS screening in serum samples following a simple methyl-tert-butyl ether (MTBE) liquid extraction. Separation and ionization conditions were optimized to quantify low nanomolar concentration levels of perfluorooctanoic acid (PFOA) and perfluorooctanesulfonic acid (PFOS) from serum extracts when using multiple reaction monitoring under negative ion mode conditions. Multiplexed separations of PFOA and PFOS were achieved with excellent throughput (< 3 min/sample), adequate concentration sensitivity (LOD ~ 20 nM, SNR= 3) and good technical precision over three consecutive days of analysis (mean CV = 9.1%, n = 84). Accurate quantification of PFASs was demonstrated in maternal serum samples (n = 16) when using MSI-CE-MS/MS following pre-column sample enrichment with median concentrations of 3.46 nM (0.7 to 9.0 nM) and 3.29 nM (1.5 to 6.6 nM) for PFOA and PFOS, respectively. This was lower than average PFAS exposures assessed for pregnant women who had serum collected prior to 2009 due to subsequent phase out of their production. Overall, this method offers a convenient approach for large-scale biomonitoring of environmental exposures to legacy PFASs and their emerging replacements that is relevant to maternal health and chronic disease risk in children.

"Discovering Metabolomic and Lipidomic Alterations Mediated by Genetic Mutations using an Orbitrap Exploris 480 Mass Spectrometer." (95)


Eric D. Tague, Thermo Fisher Scientific (Primary Presenter)

Sven Hackbusch, Thermo Fisher Scientific


Metabolomic experiments are hypothesis generating and aimed at attempting to identifying as many small molecules present in a system as possible during a single analysis, then comparing those compounds across varying sample groups. These types of experiments create large quantities of multi-dimensional spectral data which require advanced software to search and extract useful information.

Seven unique strains Escherichia Coli was mutated at the gene level to create robust and reproducible biological variants, each lacking a different enzyme of the central carbon metabolism. Metabolites were extracted using chilled acidic acetonitrile and methanol solution before being separated on a Thermo Scientific™ Hypersil GOLD™ column. Lipids were extracted using a mixture of ethanol, water, diethyl ether, pyridine and ammonium hydroxide prior to separation on a Thermo Scientific™ Accucore™ C18 + column. Mass spectrometric data were collected using the Thermo Scientific™ Orbitrap Exploris™ 480 mass spectrometer incorporating features like EASY-IC™ and optimized dd-MS2 experiments. Data files were analyzed with Thermo Scientific™ Compound Discoverer™ 3.1 for metabolomics and LipidSearch™ 4.2 for lipidomics data utilizing new modules for normalization and quantitation.

For the metabolomics samples, the acquired data was normalized on a per compound basis using a pooled QC sample, which was injected every 10 samples, and scaled to Optical Density measurements within Compound Discoverer. Additionally, system suitability could be confirmed by plotting the area counts of an added internal standard. Excellent reproducibility was seen with CVs below 4% in both polarities. Using uni- and multivariate analyses like volcano plots and principle component analyses, the acnB mutant was found to exhibit a distinct metabolomic profile compared to all other mutants and the WT strain. No mutant stains showed excessive oxidative stress when compared to the WT, based on relative abundances of glutathione and glutathione between the samples. Metabolika pathway mapping was used to visualize abundance changes for compounds in the TCA cycle. There was a large fold change for all compounds that exist upstream of the acnB mutation but not for acnA, signifying these two enzymes don’t function under the same cellular conditions.

After performing peak detection and alignment in LipidSearch 4.2™, the major lipid classes present in the E. coli mutants were found to be phosphatidylglycerol and phosphatidylethanolamine. This is consistent with the known membrane composition of this bacterium. While there were a number of triglyceride and phosphatidylcholine lipids identified in the samples, their origin could easily be determined to not be culture dependent with the use of media blank controls. Quantitative lipid data was generated by spiking an internal standard mixture (Avanti® SPLASH® LIPIDOMIX®) into all samples, containing one labeled lipid per common compound class. It was seen that the top 10 most common PG compounds in bacterial cells are more abundant in the acnB mutant stain, suggesting substantial metabolic diversion to lipid biosynthesis and potential cell membrane anomalies.

"A Comprehensive Targeted Metabolomic Assay for Uremic Toxin Quantification" (111)


Lun Zhang, University of Alberta (Primary Presenter)

Jiamin Zheng, University of Alberta

Maheswor Gautam, University of Alberta

Rupasri Mandal, University of Alberta

David Wishart, University of Alberta


As the global population ages, kidney failure, or reduced kidney function is becoming a major burden to many health-care systems. Kidney failure or reduced kidney function leads to a reduced Glomerular Filtration Rate (GFR). This, in turn, leads to the accumulation of endogenously produced toxic solutes in the blood and the development of a condition known as uremia. These toxic solutes are called uremic toxins. Uremic toxins include dozens of well-known small molecules such as urea, hippuric acid, uric acid and creatinine. Many uremic toxins are actually the byproducts of gut metabolism (such as TMAO, bile acids, cresol sulfate and indoxyl sulfate) and there appears to be a close link between diet, gut microflora and circulating levels of certain uremic toxins. At low abundance, most uremic toxins are relatively harmless, however at high levels, uremic toxins can contribute to diabetes, liver failure, heart disease, memory loss, kidney disease, jaundice and a variety of intractable of skin conditions. Indeed, it is believed that uremic toxins contribute to many of the chronic conditions seen in the elderly – even those without overt kidney disease. Given the importance of uremic toxins in health and disease, especially for those with kidney disease and those with reduced kidney function, we believe that an improved method for detecting and quantifying uremic toxins is needed. To date, the methods published for uremic toxin characterization are only able to identify and quantify a handful of uremic toxins. Here we describe a new assay based on liquid chromatography-tandem mass spectrometry (LC-MS/MS) that allows us to identify and quantify 89 small-molecule uremic toxins. This includes 35 organic acids, 31 amino acids and derivatives, 14 biogenic amines and derivatives, and 9 nucleotides/nucleosides. In developing the assay we carefully checked the accuracy, reproducibility, sensitivity and inter-day stability. The accuracies of quality control solutions and the recovery rates of spiked human pooled-serum samples at 3 different concentrations were in range of 100 ± 10% and 100 ± 20%, respectively. As far as we are aware, this is the most comprehensive, fully quantitative metabolomic assay that has been reported for the analysis of uremic toxins. We further demonstrate the utility of this uremic toxin assay by reporting the levels of many previously unmeasured and potentially important uremic toxins in the serum samples of patients undergoing renal dialysis.