Fig. 1: The flow chart of our pipeline.
From: Non-coding genetic variants underlying higher prostate cancer risk in men of African ancestry
Non-coding SNPs in 1000 Genomes Project are retained if they are common in either EA or AA cohort, and have a high allele frequency difference between the two populations. The deep learning model integrated with a sliding window strategy was then applied to the retained 491 K SNPs to select the ~2000 AA-dominant SNPs (eSNPs) at which the model predicts substantially different enhancer activities for the two alleles, using a criterion called essential window number (EWN) that measures consistency of the directional impact of functional SNPs on enhancer activity (Methods). The eSNPs are categorized as either gained, where the alternate allele has greater predicted enhancer activity, or lost, where the alternate allele has lower predicted enhancer activity. These ~2000 eSNPs are used to assess the polygenic risk score (PRS) for PrCa for an individual. Part of the figures are generated by BioRender.
