Identifying Transferrable Poly-Omics Risk Scores for Suicide Attempt in a Longitudinal Diverse Cohort

Suicide attempts (SA) has been estimated to occur about 10-20 times more frequently than actual suicide death.  Existing SA prediction models usually leverage electronic health records only, and lack of consideration for genetic vulnerability and molecular risk factors. Without biological mechanism support, it is hard to generalize and transfer risk models across different race/ethnicity populations.  However, previous genetics research that is dominated by European-descent populations also exacerbate health disparity; in particular, African Americans are much less represented.  With the rapid accumulation of longitudinal population-based cohorts, large-scale genomic data for multi-ethnic groups, and updated psychiatric genomics resources, we can now integrate different datasets together to identify omics-based biomarkers and risk scores that may better associate with SA.  We hypothesize that SA risk can be better predicted or stratified when considering genomic, transcriptomic, and proteomic risk scores together.  In addition, we believe higher-level functional units (genes and pathways) are more transferrable across different ancestries, hence should be incorporated to increase portability of the trans-ethnic risk prediction models.  The overall goal is to identify the most relevant omics predictors in association with SA, and then build and validate predictive risk models for SA onset among both European-descent and African-descent populations.  To achieve it, we have been approved data access to raw diverse cohorts in All of Us research program, genome-wide association summary statistics from International Suicide Genetics Consortium and Million Veteran Program, and functional genomics data from GTEx and PsychEncode project.  This contributes to poly-Omics scoring of >8,000 SA cases (46% non-EUR) and >80,000 matched non-SA controls. The first aim is to identify individual risk and resilient factors for SA through genetic epidemiology and integrative omics analysis in All of Us retrospective cohort. In the second aim, we will evaluate the clinical utility of the poly-Omics scoring based risk model for SA development in All of Us prospective cohort. Optimized model will better classify individuals at low, moderate, and high risk for developing SA. Together these scores and models will allow us to develop new strategy and approach for future SA screening and prevention.  This project will be the first ever to use all these unique and generalizable data sources together to evaluate the clinical value of genomics data for suicide prevention.  For next step, we will seek additional replication in other population-based cohorts (e.g., UK Biobank and Kaiser Permanente Research Bank).  We will also leverage the dissemination framework of Henry Ford Health and Mental Health Research Network to quickly implement this new model once its incremental value is confirmed in this project. The framework will also be able to generalize for studies of other mental health conditions for the purpose of better risk characterization.