We analyze the same J = 394,174 SNPs and G = 18, 364 SNP-sets used in the Framingham Heart Study analyses. Here, SNP-set annotations are based on gene boundaries defined by the NCBI’s RefSeq database in the UCSC Genome Browser [50]. Unannotated SNPs located within the same genomic region were labeled as being within the “intergenic region” between two genes. This file gives the posterior inclusion probabilities (PIPs) for the input and hidden layer neural network weights after fitting the BANNs model on the individual-level data. We assess significance for both SNPs and SNP-sets according to the “median probability model” threshold 57. Page #1 provides the variant-level association mapping results with columns corresponding to: (1) chromosome; (2) SNP ID; (3) chromosomal position in base-pair (bp) coordinates; (4) SNP PIP; and (5) SuSiE PIP, which corresponds to SNP-level posterior inclusion probabilities computed by SuSiE [46]. Page #2 provides the SNP-set level enrichment results with columns corresponding to: (1) chromosome; (2) SNP-set ID; (3-4) the starting and ending position of the SNP-set chromosomal boundaries; (5) SNP-set PIP; (6) RSS PIP, which corresponds to the posterior inclusion probabilities computed by RSS [26]; (7) the number of SNPs that have been annotated within each SNP-set; (8) the “top” associated SNP within each SNP-set; (9) the PIP of each top SNP. Pages #3 and #4 provide similar results based on analyses where each SNP-set annotation has been augmented with a ±500 kilobase (kb) buffer to account for possible regulatory elements. (ZIP)

Please read README.md for more information. We also provide the iPGS browser (https://ipgs.mit.edu/), where you can browse and download the coefficients of our iPGS models.