Understanding and predicting the functional consequences of missense mutations in BRCA1 and BRCA2

Abstract

BRCA1 and BRCA2 are tumour suppressor genes that play a critical role in maintaining genomic stability via the DNA repair mechanism. DNA repair defects caused by BRCA1 and BRCA2 missense variants increase the risk of developing breast and ovarian cancers. Accurate identification of these variants becomes clinically relevant, as means to guide personalized patient management and early detection. Next-generation sequencing efforts have significantly increased data availability but also the discovery of variants of uncertain significance that need interpretation. Experimental approaches used to measure the molecular consequences of these variants, however, are usually costly and time-consuming. Therefore, computational tools have emerged as faster alternatives for assisting in the interpretation of the clinical significance of newly discovered variants. To better understand and predict variant pathogenicity in BRCA1 and BRCA2, various machine learning algorithms have been proposed, however presented limited performance. Here we present BRCA1 and BRCA2 gene-specific models and a generic model for quantifying the functional impacts of single-point missense variants in these genes. Across tenfold cross-validation, our final models achieved a Matthew's Correlation Coefficient (MCC) of up to 0.98 and comparable performance of up to 0.89 across independent, non-redundant blind tests, outperforming alternative approaches. We believe our predictive tool will be a valuable resource for providing insights into understanding and interpreting the functional consequences of missense variants in these genes and as a tool for guiding the interpretation of newly discovered variants and prioritizing mutations for experimental validation.

Identifying the Molecular Drivers of Pathogenic Aldehyde Dehydrogenase Missense Mutations in Cancer and Non-Cancer Diseases

“…(b) Supervised machine learning Supervised machine learning algorithms were used to investigate the link between the biochemical and functional annotations of mutations with pathogenic and benign labels. We applied the same analysis pipeline to this work, which has been successfully used in the characterisation of the effects of mutation on other proteins [16][17][18].…”

“…This complexity has hindered efforts to rationally and systematically characterise the role of ALDH mutations in diseases. We have previously shown that computational tools can be used to understand the consequences of missense mutations on protein structure, providing insight into molecular mechanisms of disease and further predicting disease outcome [16][17][18]. Towards better understanding the molecular consequences of diseaseassociated ALDH mutations, here, we have curated a set of high-confidence clinically The disruption of ALDH activity has been linked to a wide range of diseases including epilepsy [6,7], alcohol liver disease [8], sjorgen-larrson syndrome [9], hyperprolinemia [10], hyperammonemia [11,12], and aciduria [13].…”

A review of genetic variant databases and machine learning tools for predicting the pathogenicity of breast cancer

“…REVEL performed very well in predicting the pathogenicity of variants compared with individual tools [ 102 ]. Although REVEL was not initially developed for predicting BC pathogenic variants, it has shown good performance with an area under the curve (AUC) of 0.79, which is one of the highest accuracy values compared with tools not designed specifically for BC [ 103 ].…”

Gene-specific machine learning for pathogenicity prediction of rare BRCA1 and BRCA2 missense variants

“…In this sense, it is likely to perform poorly due to high variance. Crockett et al 23 , Padilla et al 24 , Hart et al 21 , and Aljarf et al 19 have developed gene-speci c variant pathogenicity predictors for disease-associated genes, including BRCA1 and BRCA2. Their studies have shown that gene-speci c predictors perform better than or comparably to genome-wide ones.…”

“…Supervised machine learning has been widely adopted to develop computational tools for the pathogenicity prediction of variants, including rare missense ones [13][14][15][16][17][18][19][20][21] . A prediction tool based on supervised machine learning takes a set of features, such as minor allele frequencies (MAFs), predicted functional impacts of a variant, and the degree of conservation across multiple species at its genomic position, as input.…”