Genome-wide association studies (GWAS) have transformed our understanding of complex traits but are often limited by a high multiple-testing burden and the inclusion of many non-functional, non-coding variants. Our method seeks to enhance GWAS power by pre-filtering SNPs based on their predicted impact on chromatin accessibility, a key indicator of regulatory function. Utilizing the deep learning capabilities of the Enformer model, we compute SNP Activity Difference (SAD) scores that measure the effect of individual genetic variants on chromatin openness, thereby prioritizing SNPs that are likely to influence gene expression. This approach addresses the critical research question of whether integrating functional annotations can improve the recovery of true genetic associations and reveal novel loci that might be missed by traditional GWAS methods. Our objective is to systematically reduce the number of independent tests by excluding SNPs with negligible regulatory impact, allowing for a less stringent significance threshold and, ultimately, increasing the power to detect meaningful associations. By linking genetic variation directly to regulatory mechanisms, our method not only streamlines the analytical pipeline but also enhances the biological interpretability of GWAS findings, paving the way for deeper insights into the genetic architecture of complex traits.
