Fitness of unregulated human Ras mutants modeled by implementing computational mutagenesis and machine learning techniques

Functional analysis of BRCA1 RING domain variants: computationally derived structural data can improve upon experimental features for training predictive models

Advancements in the interpretation of variants of unknown significance are critical for improving clinical outcomes. In a recent study, massive parallel assays were used to experimentally quantify the effects of missense substitutions in the RING domain of BRCA1 on E3 ubiquitin ligase activity as well as BARD1 RING domain binding. These attributes were subsequently used for training a predictive model of homology-directed DNA repair levels for these BRCA1 variants relative to wild type, which is critical for tumor suppression. Here, relative structural changes characterizing BRCA1 variants were quantified by using an efficient and cost-free computational mutagenesis technique, and we show that these features lead to improvements in model performance. This work underscores the potential for bench researchers to gain valuable insights from computational tools, prior to implementing costly and time-consuming experiments.

Structure-based functional analysis of BRCA1 RING domain variants: Concordance of computational mutagenesis, experimental assay, and clinical data

A significant impediment to the improvement of clinical outcomes in treating breast and ovarian cancers rests with the lack of available interpretations for BRCA1 variants of unknown significance. Two research groups recently implemented large-scale functional assays for quantifying effects of single missense mutations on homology-directed DNA repair activity of BRCA1 variants, which is critical for tumor suppression and strongly correlates with cancer risk, and their results are significantly concordant with each other as well as with known pathogenic and benign variant clinical data. In this work, we implemented an established computational mutagenesis procedure to characterize structural impacts of single residue replacements to the BRCA1 RING domain. The computational data showed similarly strong concordance with known clinical data as well as with experimental data from both functional assays. Predictions made by models trained on our computational data offer a complementary and orthogonal approach for classifying all remaining unexplored BRCA1 RING domain variants.