Gulf War Illness (GWI) Metabolomic Analysis

Bioinformatics Internship Presentation

Kelvin Carrasquillo-Carrión (Mentor: Dr. Simina Boca, Innovation Center for Biomedical Informatics, Department of Oncology, Georgetown University)

September 1st, 2015, 2:20-2:40pm, Room 1300, Harris Building

More than 170,000 of the estimated 700,000 military personnel who served during the 1990-1991 Gulf War were diagnosed with Gulf War Illness (GWI).  GWI is often confused in symptomatology with Chronic Fatigue Syndrome (CFS). Recent studies describe two different phenotypes in patients diagnosed with GWI after physical activity, namely: START (Stress Test Associated Reversible Tachycardia) and STOPP (Stress Test Occurring Phantom Perception).  The goal of this study is to provide distinctive biomarkers for GWI and CFS, as well as possibly for the START and STOPP phenotypes, from cerebrospinal fluid (CSF) biospecimens. Our analysis considers adjustments for gender, BMI, and age, as these variables have shown associations with a number of metabolites in previous studies. For this project we analyzed a total of 114 patients (29 healthy individuals, 43 CFS cases, 42 GWI cases); the samples included both pre-physical and post-physical activity measurements. The post-exercise samples consisted of 7 healthy individuals and 22 GWI individuals, out of which 7 had the START phenotype and 15 the STOPP phenotype. Thus, a total of 6 groups were considered: healthy individuals before exercise, GWI individuals before exercise, CFS individuals before exercise, healthy individuals after exercise, GWI START individuals after exercise, and GWI STOPP individuals after exercise

To determine which peaks were significantly different between the 6 groups, we considered linear models which had terms for the groups, as well as for BMI, gender, and age, with the transformed and normalized peak intensities as the outcomes. For each peak, we tested whether there was an association between the intensity and the group membership. Given that there were 1217 peaks, we applied a multiple testing correction using the FDR at the 0.05 level, leading to 39 peaks which were significantly different between the groups, 17 in the positive mode and 22 in the negative mode.   A two-step annotation procedure was performed to search for the corresponding metabolites and peptides in the HMDB and METLIN databases. The annotations for the positive mode gave 9/17 peaks with metabolites not showing disease associations in previous studies, 3/17 peaks with metabolites with disease associations in previous studies, and 5/17 peaks annotated to oligopeptides. The annotations in negative mode gave 2/22 peaks annotated to metabolites and one peak annotated to oligopeptides; no metabolites showing disease associations in previous studies were found in the negative mode.  Thus, this step resulted in 6/39 peaks annotated to peptides over both modes; these oligopeptides may play a role as neurotransmitters in signaling pathways in the brain. Since the metabolite and peptide annotations result in a number of possible metabolites for the same m/z value per peak, MS/MS validation is required to determine the exact peak identities.