Protein Feature Enrichment Score (PFES): proteome-wide mechanistic partitioning of missense variant

Data repository for precomputed Protein Feature Enrichment Scores (PFES) for all possible missense variants across the human proteome (calculated on Aug 2025). Files are split by chromosome. PFES quantifies the degree to which a missense variant's protein characteristics resemble those statistically enriched among pathogenic or benign variants. For each variant, the score is computed as the sum of log odds ratios across protein features (see this glossary for description of all protein features used in this study) showing significant enrichment (FDR-corrected p < 0.01) in pathogenic versus control variants, calculated within 20 PANTHER protein functional classes. Positive PFES values indicate enrichment for pathogenic-associated protein characteristics; negative values indicate enrichment for benign-associated protein characteristics. Variants are further partitioned into three categories: PF-Enriched (statistically distinct from the benign distribution), PF-Depleted (statistically distinct from the pathogenic distribution), and PF-Neutral (consistent with both distributions) based on the empirical distributions of existing pahogenic and benign variants. Each file contains the following columns: UniProtID: UniProt accession identifier for the canonical protein isoform Gene: HGNC gene symbol Mutation: Amino acid substitution in standard single-letter notation (e.g., A123V) PFES: Overall Protein Feature Enrichment Score PFES_Partitioning: Variant category (PF-Enriched, PF-Neutral, or PF-Depleted) p_value: Statistical significance of deviation from the reference distribution used for partitioning, either p_enriched for positive PFES or p_depleted for negative PFES PFES_Physicochemical: PFES contribution from physicochemical properties of the amino acid substitution PFES_Function: PFES contribution from functional site annotations PFES_Domain: PFES contribution from domain and region annotations PFES_Modification: PFES contribution from post-translational modification sites PFES_Structure: PFES contribution from 3D structural features PFES_PPI: PFES contribution from protein-protein interaction interface features Scores were computed using protein structural annotations from PDB and AlphaFold, sequence and functional annotations from UniProtKB and PhosphoSitePlus, and protein class assignments from PANTHER (v19.0). Variant pathogenicity reference distributions were derived from 85,321 pathogenic variants (ClinVar/HGMD) and 130,719 control variants (ClinVar benign/likely benign and gnomAD common variants with AF ≥ 1%). For details on score computation, partitioning thresholds, and feature annotations, see the accompanying manuscript (DOI:) or GitHub.