Biological constraints on GWAS SNPs at suggestive significance thresholds reveal additional BMI loci

To uncover novel significant association signals (p<5×10^-8), genome-wide association studies (GWAS) requires increasingly larger sample sizes to overcome statistical correction for multiple testing. As an alternative, we aimed to identify associations among suggestive signals (5 × 10^-8≤p<5×10^-4) in increasingly powered GWAS efforts using chromatin accessibility and direct contact with gene promoters as biological constraints. We conducted retrospective analyses of three GIANT BMI GWAS efforts using ATAC-seq and promoter-focused Capture C data from human adipocytes and embryonic stem cell (ESC)-derived hypothalamic-like neurons. This approach, with its extremely low false-positive rate, identified 15 loci at p<5×10^-5 in the 2010 GWAS, of which 13 achieved genome-wide significance by 2018, including at NAV1, MTIF3, and ADCY3. Eighty percent of constrained 2015 loci achieved genome-wide significance in 2018. We observed similar results in waist-to-hip ratio analyses. In conclusion, biological constraints on sub-significant GWAS signals can reveal potentially true-positive loci for further investigation in existing data sets without increasing sample size.