6. GENOME-WIDE ASSOCIATION STUDIES OF PSYCHOTIC DISORDERS IN DIVERSE AFRICAN POPULATIONS REVEAL SHARED AND POPULATION-SPECIFIC GENETIC RISK

Background Psychotic disorders are highly heritable severe mental illnesses with suboptimal treatment outcomes for many affected individuals. Genetics research offers a critical avenue for understanding disease mechanism, yet African populations remain substantially underrepresented in psychiatric genetics despite their extensive genetic diversity. Methods We conducted genome-wide association studies (GWAS) in the Neuropsychiatric Genetics in African Populations (NeuroGAP) cohort—the largest psychiatric genetics study of its kind in Africa—including 10,919 schizophrenia cases, 5,573 bipolar disorder cases, and 16,176 controls recruited across five sites in Ethiopia, South Africa, Uganda, and Kenya. Using mega-analysis, meta-analysis, and cross-ancestry comparative approaches, we evaluated genetic associations within NeuroGAP and characterized shared and population-specific genetic architecture relative to Psychiatric Genomics Consortium (PGC) datasets. Results We identified 10 suggestive loci (P < 1e-6) in the mega-analysis that are enriched in African populations compared to others including rs112832881 near DCKL1 (P=7.694E-07, SE=0.06) and rs16870956 near ARHGEF28 (P=6.085E-07,SE=0.05). Site-specific analyses further revealed seven genome-wide significant loci (P < 5-e8) including rs113085686 near RACGAP1 (P_KEMRI=1.66E-08, SE=0.18) and rs112715845 near CEP192 (P_South Africa=4.62E-08, SE=0.19); underscoring substantial within-Africa genetic heterogeneity. Despite broadly consistent genetic architectures indicated by strong cross-ancestry genetic correlations between PGC and NeuroGAP (r₉=0.72 ± 0.08 for schizophrenia and r₉=0.84 ± 0.14 for bipolar disorder), within-NeuroGAP genetic correlations were vastly variable – ranging from 0.3-0.7 for schizophrenia and 0.08-0.89 for bipolar disorder. A meta-analysis between NeuroGAP and PGC refined known loci where effects were directionally consistent, while differences in allele frequency and effect size attenuated association signals. Polygenic risk scores derived from PGC showed variable predictive performance across sites, with the best performing PRS in Ethiopia (R2liability=0.029, P=7.38E-12) and the worst performing in Uganda (R2liability=0.0082, P=7.14e-24) reflecting population divergence. Multi-ancestry PRS outperformed single ancestry PRS ranging from 0.005-0.04, but was still markedly lower than the best performing PRS in European-ancestry populations (R2∼0.07). Discussion These findings demonstrate both shared and population-specific genetic risk for psychotic disorders and emphasize the importance of expanding genomic studies in diverse populations to improve global understanding of the biology of psychiatric disorders.