Genetic study of autism in African Americans shows limited transferability of European-derived polygenic scores

The absence of non-European cohorts in genetic studies of neurodevelopmental and neuropsychiatric disorders severely limits the understanding of their full genetic architecture and undermines implementation of precision medicine. Here, we directly addressed this issue by recruiting African Americans (AfrAms) with autism spectrum disorder (ASD) and analyzing their rare and common genetic variation. We performed both global and local ancestry analyses to characterize the complex patterns of admixture at the individual level and compare genetic factors between European (EUR) and African (AFR) genetically inferred ancestries (GIAs) across multiple cohorts in a total of 38,483 autistic individuals. We showed consistent common variant genetic effect sizes for ASD in EUR and AFR GIAs through genome-wide association studies. We demonstrated the limited transferability of EUR-derived polygenic scores (PGSs) based on polygenic transmission disequilibrium and ancestry partial PGS analysis. We found significant autism association for high-impact rare copy number variants in both GIAs. We identified a set of candidate ASD loci based on rare deletions observed in AFR GIA carriers, including SMC2, DMTN, SORCS1, and ROGDI, and detected a signal for de novo missense variants of predicted low impact in AFR GIA individuals. Finally, we uncovered significant depletion of AFR GIA autistic carriers of rare variants in known associated genes found in EUR cohort studies. These findings are the first to detail common and rare variant genetic contributions to ASD in AfrAms and demonstrate that their involvement in neurodevelopmental and neuropsychiatric disorders' genomic research is essential to advance discovery.