Abstract
<title>Abstract</title> <p> Background Breast Cancer 1 ( <italic>BRCA1</italic> ) gene variations are known to be associated with breast and ovarian cancers. At the same time, interpreting <italic>BRCA1</italic> missense variants remains challenging, particularly for variants of uncertain significance (VUS), because sequence-based predictors do not fully capture local structural context. AlphaFold-derived predicted Local Distance Difference Test (pLDDT) scores provide residue-level confidence estimates that distinguish well-ordered regions from regions of lower structural confidence. We assessed whether local pLDDT is associated with the clinical classification of BRCA1 missense single-nucleotide variants and whether it stratifies the performance of commonly used in silico predictors. Results BRCA1 missense variants collected in the ClinVar database were filtered, annotated with the Ensembl Variant Effect Predictor, mapped to residue-level pLDDT values from the AlphaFold BRCA1 model, and integrated with UniProt domain annotations. The final dataset comprised 2,399 unique missense variants, 2,390 of which had mappable pLDDT values. Pathogenic and likely pathogenic variants were strongly enriched in regions with pLDDT > = 70 relative to benign and likely benign variants (odds ratio 30.13, 95% confidence interval 17.95–50.60; Fisher’s exact test p = 1.41 × 10⁻⁶¹). This association remained significant within the BRCT region (odds ratio 9.35, 95% confidence interval 2.04–42.93; p = 0.00109) and in ClinVar subsets with stronger review support. Continuous pLDDT distributions also differed markedly between pathogenic/likely pathogenic and benign/likely benign variants (Mann–Whitney U test p = 1.31 × 10⁻⁶⁰). In receiver operating characteristic analysis, SIFT outperformed PolyPhen overall (AUC 0.872 vs 0.692), whereas predictor performance differed across pLDDT strata. Conclusions In BRCA1, AlphaFold-derived local confidence was strongly associated with clinical missense variant classification and may serve as useful contextual structural evidence rather than as a standalone pathogenicity score. pLDDT may therefore help refine variant prioritization and support the interpretation of BRCA1 VUS in conjunction with existing computational and clinical evidence. </p>