Knowledge-Assisted Multi-omic Biomarker Identification for Posttraumatic Stress Disorder
Abstract
There is a pressing need to identify reliable molecular and physiological biomarkers of posttraumatic stress disorder (PTSD) for the accurate diagnosis, prognosis, and treatment of the disorder. The DOD-funded Systems Biology of PTSD Consortium has recruited over 200 male combat veterans with and without PTSD for the purposes of identifying these biomarkers. Whole blood samples taken from each subject have been used to isolate DNA, RNA, protein, metabolites, and endocrine markers for subsequent study. During the first phase of analysis, candidate biomarkers for PTSD were proposed using data from each of the measured modalities. However, upon coupling these biomarkers with supervised machine learning techniques to predict the PTSD status of new subjects, the resulting classifiers achieved only modest performance. The overall goal of this proposal is to improve the sensitivity and specificity of candidate PTSD biomarkers through knowledge-driven and knowledge-guided methods for data integration and classification. Knowledge-driven methods will involve manually and algorithmically selecting metabolic and signaling pathways relevant to PTSD from training datasets. Constituent biomolecules (e.g., mRNAs, metabolites, proteins) from these pathways will then be interrogated for biomarkers using customized feature selection and supervised classification techniques. As a complementary approach, knowledge-guided methods will involve the development and application of computationally efficient probabilistic models for incorporating genome-wide regulatory relationships (e .g., transcription factor, DNA methylation, and microRNA interactions) with multi-omic data. These relationships, derived from both prior biological knowledge and inferred directly from PTSD data, will enable discovery of network-level biomarkers that provide much-needed mechanistic insight into PTSD. In addition to these approaches, collaborative work outlined in this proposal will explore PTSD subgroup-specific molecular mechanisms and biomarkers. To assist in this effort, members of the Consortium have (1) re-collected samples from 70 male veterans at two additional time points and (2) recruited 80 female veterans and approximately 1700 active duty soldiers . Proposed analyses will examine changes in male veteran subgroup membership over time as well as identify subgroup differences between veteran males, veteran females, and active duty personnel. In all cases, candidate PTSD biomarkers will be evaluated using supervised classification performance measures such as the area under the receiver operating characteristic curve (AUC). Proposal deliverables will include a deeper understanding of molecular mechanisms underlying PTSD, lists of candidate biomarkers, estimated classification performance, and novel computational tools which will be readily applicable to other conditions of strategic importance to the Army.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 07, 2018
- Source ID
- W911NF1710069
Entities
People
- Bernie J Daigle
Organizations
- Army Contracting Command
- United States Army
- University of Memphis