NAc reduced food seeking and increased avoidance re- sponses. In a slice recording preparation, photoactivation of aPVTCRF fibers in the NAc elicited large excitatory post- synaptic responses that were blocked with glutamatergic antagonists. Conclusions: Together our results suggest that the aPVTCRF-NAc pathway regulates the balance between avoiding threats and searching for rewards. Supported By: NIH R00-MH105549; Rising STARs Award from University of Texas System; NARSAD Young Investigator Keywords: Paraventricular Thalamus, Nucleus Accumbens, Corticotropin-Releasing Factor, Predator Odor, Food Seeking SYMPOSIUM Subtypes of PTSD in Military Veterans Based on Biological and Clinical Variables Chair: Charles Marmar Co-Chair: Marti Jett Identifying Subtypes of PTSD Carole Siegel 1 , Eugene Laska 1 , Ziqiang Lin 1 , and Charles Marmar 2 1 New York University School of Medicine, 2 NYU Langone Health Background: The purpose is to identify subtypes of PTSD for males post military service from clinical assessments only. We investigate whether biological and risk factors can distinguish those without PTSD from each of the identified subtypes. Predictive biological features would give substantial evidence of the validity of the subtypes and help elucidate the role of these features in the treatment and biology of PTSD. Viewing PTSD as heterogeneous by searching for subtypes might help elucidate the biology of resilience. Methods: A machine learning method, Random Forests (RF) with Partitioning around Medoids, (RF/PAM) was used to identify clinical subtypes. Validation of subtypes included internal RF methods (bootstrapping), and external validation by separate RF confirmatory association with biological and other potential prediction features of group membership. The sample size is 145, 71 normal and 74 with clinically diagnosed PTSD. Results: Severity of symptoms distinguishes two PTSD subtypes (S1, n¼ 27; S2 , n¼47). They differed by severity on every one of the defining clinical assessments (S2>S1) with S2 having more severe scores on all assessments. S1 assessment scores were less coherent and had some overlap with normals. Normals and subtypes statistically differed (p<.001) on values of methylation variables, neurocognition scores, having lifetime depression and past trauma. RF results yielded AUCs for ROCs for both subtyping and classification of greater than .85 indicating excellent model fits. Conclusions: Differential treatment approaches for the subtypes as well as different biological causation for the more severe cases is suggested by these findings. Supported By: US Army, Department of Defense Keywords: Complex PTSD, DNA Methylation, Neuro- cognition, Machine Learning Epigenetic Biotypes of PTSD in War-Zone Exposed Veteran and Active Duty Males Ruoting Yang 1 , Aarti Gautam 2 , Derese Getnet 1 , Bernie Daigle 3 , Stacy Ann Miller 4 , Kelsey Dean 5 , Seid Muhie 6 , Kai Wang 7 , Inyoul Lee 7 , Duna Abu Amara 8 , Janine D. Flory 9 , PTSD Systems Biology Consortium, Leroy Hood 7 , Owen Wolkowitz 10 , Synthia Mellon 10 , Francis J. Doyle III 5 , Rachel Yehuda 9 , Charles Marmar 11 , Kerry Ressler 12 , Rasha Hammamieh 2 , and Marti Jett 2 1 Walter Reed Army Institute of Research, 2 U.S. Army Center for Environmental Health Research, 3 The University of Memphis, 4 ORISE, 5 Harvard University, 6 USACEHR, 7 Institute for Systems Biology, 8 Steven and Alexandra Cohen Veterans Center for the Study of Posttraumatic Stress and Traumatic Brain Injury, NYU School of Medicine, 9 James J. Peters Veterans Administration Medical Center Bronx/Icahn School of Medicine at Mount Sinai, 10 Univer- sity of California, San Francisco, 11 NYU School of Medi- cine, 12 McLean Hospital Background: That PTSD is a heterogeneous condition is supported by both the failure to identify objective physiological measurements applicable to all who meet criteria for the disorder, and divergent responses associated with PTSD treatments. Methods: This study attempted to capitalize on biological di- versity observed following epigenome-wide analysis in a well- characterized male veteran Discovery cohort (N¼166) consist- ing of 83 PTSD+ and 83 PTSD- participants, to identify biolog- ically relevant PTSD subtypes (biotypes) that might further improve molecular diagnosis and personalized treatment. Initial analysis revealed associations between DNA methylation (DNAm) profiles and 34 clinical features from which two epige- netically distinct biotypes e G1 and G2 e were derived. Results: The findings were validated by examining partici- pants (N¼59) at a 3-year follow-up. Two other independent, cross-sectional veteran cohorts (N¼54, and N¼38, respec- tively), and a longitudinal active duty cohort (N ¼ 133) were also used for validation of the initial biotypes. Interestingly, the biological pathways associated with the biotypes appeared to be regulated in opposite directions in comparison to controls. The impact of biotype-specific signal were evaluated in pub- lished DNAm markers, including an independent multi-omics biomarker developed using the same veteran cohort. Finally, we demonstrated filtering biotype-specific signal from a prior marker would result in a high-performance marker (AUC of 0.85). Conclusions: The identification of two novel distinct epige- netic biotypes for PTSD may have future utility in under- standing biological and clinical heterogeneity in PTSD and potential applications in treatment-matching and monitoring of clinical outcome. Supported By: This work was supported by funding from the U.S. Army Research Office, through award S8 Biological Psychiatry May 1, 2020; 87:S1eS105 www.sobp.org/journal Symposium Abstracts Biological Psychiatry