Electronic Theses and Dissertations



Document Type


Degree Name

Doctor of Philosophy



Committee Chair

Bernie Daigle, Jr.

Committee Member

Ramin Homayouni

Committee Member

Jennifer Mandel

Committee Member

Deepak Venugopal


Post-traumatic stress disorder (PTSD) is a psychiatric disorder caused by environmental and genetic factors resulting from alterations in genetic variation, epigenetic changes and neuroimaging characteristics. There is a pressing need to identify reliable molecular and physiological biomarkers for accurate diagnosis, prognosis, and treatment, as well to deepen the understanding of PTSD pathophysiology. Machine learning methods are widely used to infer patterns from biological data, identify biomarkers, and make predictions. The objective of this research is to apply machine learning methods for the accurate classification of human diseases from genome-scale datasets, focusing primarily on PTSD.The DoD-funded Systems Biology of PTSD Consortium has recruited combat veterans with and without PTSD for measurement of molecular and physiological data from blood or urine samples with the goal of identifying accurate and specific PTSD biomarkers. As a member of the Consortium with access to these PTSD multiple omics datasets, we first completed a project titled Clinical Subgroup-Specific PTSD Classification and Biomarker Discovery. We applied machine learning approaches to these data to build classification models consisting of molecular and clinical features to predict PTSD status. We also identified candidate biomarkers for diagnosis, which improves our understanding of PTSD pathogenesis. In a second project, entitled Multi-Omic PTSD Subgroup Identification and Clinical Characterization, we applied methods for integrating multiple omics datasets to investigate the complex, multivariate nature of the biological systems underlying PTSD. We identified an optimal 2 PTSD subgroups using two different machine learning approaches from 82 PTSD positive samples, and we found that the subgroups exhibited different remitting behavior as inferred from subjects recalled at a later time point. The results from our association, differential expression, and classification analyses demonstrated the distinct clinical and molecular features characterizing these subgroups.Taken together, our work has advanced our understanding of PTSD biomarkers and subgroups through the use of machine learning approaches. Results from our work should strongly contribute to the precise diagnosis and eventual treatment of PTSD, as well as other diseases. Future work will involve continuing to leverage these results to enable precision medicine for PTSD.


Data is provided by the student.

Library Comment

Dissertation or thesis originally submitted to ProQuest