Comprehensive annotation of BRCA1 and BRCA2 missense variants by functionally validated sequence-based computational prediction models

Abstract

Purpose: To improve methods for predicting the impact of missense variants of uncertain significance (VUS) in BRCA1 and BRCA2 on protein function.

Methods: Functional data for 248 BRCA1 and 207 BRCA2 variants from assays with established high sensitivity and specificity for damaging variants were used to recalibrate 40 in silico algorithms predicting the impact of variants on protein activity. Additional random forest (RF) and naïve voting method (NVM) metapredictors for both BRCA1 and BRCA2 were developed to increase predictive accuracy.

Results: Optimized thresholds for in silico prediction models significantly improved the accuracy of predicted functional effects for BRCA1 and BRCA2 variants. In addition, new BRCA1-RF and BRCA2-RF metapredictors showed area under the curve (AUC) values of 0.92 (95% confidence interval [CI]: 0.88-0.96) and 0.90 (95% CI: 0.84-0.95), respectively. Similarly, the BRCA1-NVM and BRCA2-NVM models had AUCs of 0.93 and 0.90. The RF and NVM models were used to predict the pathogenicity of all possible missense variants in BRCA1 and BRCA2.

Conclusion: The recalibrated algorithms and new metapredictors significantly improved upon current models for predicting the impact of variants in cancer risk-associated domains of BRCA1 and BRCA2. Prediction of the functional impact of all possible variants in BRCA1 and BRCA2 provides important information about the clinical relevance of variants in these genes.

Keywords: BRCA1 and BRCA2; Functional evaluation; In silico prediction; Metapredictor; VUS.

Figure 1.. HDR activity of 207 BRCA2 missense variants.

The model-based HDR fold change with standard error (SE) is displayed on a logarithmic scale. The SE is included as a measure of the reproducibility of the HDR assay for each variant. Solid lines represent 99% probability of pathogenicity and 99% probability of neutrality (fold increase in GFP (+) cells < 1.66 for damaging and fold increase in GFP (+) cells > 2.41 for neutral). Dotted lines separate variants classified as deleterious, indeterminate, and neutral.

Figure 2.. Matthews Correlation Coefficients (MCC) for 42 in silico predictors with optimized thresholds for damaging versus indeterminate/neutral variants in BRCA1 and BRCA2.

Higher values indicate increased classifier performance.

Figure 3.. Estimates of the proportion of damaging missense variants by position in each gene.

The AAPOS x-axis represents the amino acid position, and the y-axis is the probability of a missense mutation being damaging from the NVM model. The lines were smoothed using a 50 amino acid sliding window.