Comprehensive molecular impact mapping of common and rare variants at GWAS loci

DNACipher predicts the effects of genetic variants across 38,582 cell type/assay combinations, over six times more contexts than previous models, enabling comprehensive variant impact analysis.

• Why it matters: Understanding the molecular effects of variants at GWAS loci is crucial for elucidating their roles in complex traits, but existing models are limited to specific cell types and assays.
• What they did: The study developed DNACipher, a deep learning model that takes 524 kb of genome sequence to predict variant effects across diverse biological contexts, and introduced DVIM to prioritize variants based on predicted molecular effects.
• The result: Applying DVIM to 590,604 variants at GWAS loci across 10 traits, it identified imVars at over 90% of loci for 9 traits, with common imVars aligning with fine-mapping and rare imVars enriched for trait associations, supporting their functional relevance.