Flexibly encoded GWAS identifies novel nonadditive SNPs in individuals of African and European ancestry

Most genome-wide association studies (GWAS) assume an additive inheritance model, which assigns heterozygous genotypes half the risk of homozygous-alternate genotypes. This has led to a focus on additive genetic effects in complex disease research. Growing evidence indicates that many single-nucleotide polymorphisms (SNPs) have nonadditive effects, including dominant and recessive effects, which are missed by the additive model alone. To address this issue, we developed Elastic Data-Driven Encoding (EDGE) to determine the inheritance model each SNP contributes to a given trait, allowing for unique and flexible SNP encoding in GWAS. Simulation results demonstrate that EDGE provides higher power than additive and other genetic encoding models across a wide range of simulated inheritance patterns while maintaining a conserved false positive rate. EDGE GWAS on data from the UK BioBank and the Million Veteran Program, comprising more than 500,000 individuals, identified nonadditive inheritance patterns for more than 52% of the genome-wide significant loci for coronary artery disease and body mass index. This research lays the groundwork for integrating nonadditive genetic effects into GWAS workflows to identify novel disease-risk SNPs, which may ultimately improve polygenic risk prediction in diverse populations and provide a springboard for future applications to thousands of disease phenotypes.