Skip to main content
. 2026 Oct 3;17:11324. doi: 10.1038/s41467-026-77941-y

Fig. 2. A variant-to-function heatmap of RAD51D.

Fig. 2

a Distributions of function scores by variant class; nonsense variants in codons 1–304 are perfectly separated from synonymous variants. Vertical lines denote cutoffs for function classifications (LOF, loss-of-function; INT, intermediate; NEU, neutral). Source data are provided in Supplementary Data 3. b Per-residue mean missense function scores, grouped by residue surface area exposure status and secondary structure. Unique LOF exposed residues are labeled. c Variant-to-function heatmap across RAD51D, shaded by function score (white, WT-like; red, null-like) for each mutant amino acid (rows) at each codon position (columns). Variants that do not pass multiple testing correction (lfsr>0.01) are shaded from white to gray, WT residues are boxed, and dark gray denotes missing data. Tracks above heatmaps, from top to bottom: conservation score, protein secondary structure, key domains. Asp (D) or Glu (E) are shown in boldface to highlight stronger effects of these substitutions at some sites. d Function scores separate pathogenic from benign variants among single-residue variants reported in ClinVar, and provide evidence for missense VUS and those with conflicting reports (all variants plotted SpliceAI score <0.2). Point color denotes variant type, and filled/open points denote statistical significance (i.e., filled: lfsr ≤0.01). e Precision-recall curve showing classification performance of function scores to distinguish between clinical variants with consensus interpretations of P/LP versus B/LB (n = 211 distinct variants at the protein level). f Enrichment of RAD51D LOF mutations among tumor variants in Project GENIE. Each point represents one cancer type, with P-value plotted vs odds-ratio for Fisher’s exact tests of LOF enrichment among MAVE-scored RAD51D SNVs found in that respective cancer type. Orange points: all variants tested; blue: missense (MIS) only. Dashed line denotes Bonferroni-corrected threshold for testing 11 cancers at P < 0.05. (OVCA: ovarian cancer).