Abstract
Admixed American (AMR) and Latino ancestry populations are among the least well represented groups in psychiatric genetics research. In particular, bipolar disorder has lagged behind schizophrenia and major depression in terms of the representation of non-European (EUR) populations in large-scale genome-wide association studies (GWAS).
We surveyed 52,416 Latino individuals in the Million Veteran Program (MVP) and Cooperative Studies Program (CSP) #572 for bipolar disorder and related psychosis-affective diagnoses. We generated polygenic risk scores (PRS) using PRS-CSx and performed GWAS using the linear mixed model implemented in SAIGE. Local ancestry inference was performed using RFMix2 and incorporated into association analyses using local covariate adjustment in PLINK2. We extended these PRS and GWAS analyses to 45,035 Latino individuals in the NIH All of Us Research Program (AOU) and nearly 5,000 Latino participants in the Genomic Psychiatry Cohort (GPC).
Current bipolar PRS are based on predominantly European ancestry (EUR) samples, and showed poor generalizability across ancestries. By comparison, we observed an intermediate prevalence of schizophrenia in Latinos compared to EUR and Black (AFR) participants; these rates were significantly lower among Latinos with primarily Indigenous and EUR genetic ancestry compared to Latinos of primarily EUR and AFR ancestry. Next, we performed a primary Latino GWAS of in 2,956 bipolar cases and 9,721 screened controls in MVP/CSP#572, with follow-up in 2,046 cases and 36,260 controls from AOU/GPC. We identified two genome-wide significant signals on chromosomes 15q15.3-q21.1 and 14q22.1; and six secondary associations (P < 10-6) spanning genes implicated in large-scale GWAS of psychiatric disorders, self-reported mental health problems, and lifetime educational attainment.
Our findings demonstrate that increasing ancestral diversity can drive new genomic findings. Our ongoing work extends these findings to (1) multi-ancestry meta-analysis with admixed AFR populations (10,826 cases and 64,239 controls), (2) statistical fine-mapping that leverages available GWAS summary statistics for EUR and East Asian populations, and (3) trans-diagnostic approaches based on genomic structural equation modeling.