Integrating sequence-to-omics predictions and personal multi-omics for rare variant prioritization

Watershed-DL integrates sequence-to-omics predictions with multi-omic outliers to prioritize large-effect rare variants, increasing prioritized variants by 61.4% in 1000 Genomes RNA-seq samples.

• Why it matters: Existing rare-variant prioritization methods use either sequence-based predictions or molecular outliers, leaving complementary evidence uncombined and potentially limiting diagnostic yield.
• What they did: Watershed-DL weights S2O features alongside RNA-seq outliers (n=1,030) and ATAC-seq outliers (n=174) in 1000 Genomes, and is also applied to GREGoR RNA-seq (n=677).
• The result: Prioritized variants are over 60-fold enriched for ClinVar pathogenic classifications, and matching to known pathogenic variants adds 331 recessive candidate alleles; in GREGoR, it prioritizes causal variants in 10 solved cases and finds 451 candidates in unsolved cases, including a PPP1R12A diagnostic variant and ATP5PF.