Gene-specific machine learning model to predict the pathogenicity of BRCA2 variants

Abstract

Background: Existing BRCA2-specific variant pathogenicity prediction algorithms focus on the prediction of the functional impact of a subtype of variants alone. General variant effect predictors are applicable to all subtypes, but are trained on putative benign and pathogenic variants and do not account for gene-specific information, such as hotspots of pathogenic variants. Local, gene-specific information have been shown to aid variant pathogenicity prediction; therefore, our aim was to develop a BRCA2-specific machine learning model to predict pathogenicity of all types of BRCA2 variants.Methods: We developed an XGBoost-based machine learning model to predict pathogenicity of BRCA2 variants. The model utilizes general variant information such as position, frequency, and consequence for the canonical BRCA2 transcript, as well as deleteriousness prediction scores from several tools. We trained the model on 80% of the expert reviewed variants by the Evidence-Based Network for the Interpretation of Germline Mutant Alleles (ENIGMA) consortium and tested its performance on the remaining 20%, as well as on an independent set of variants of uncertain significance with experimentally determined functional scores.Results: The novel gene-specific model predicted the pathogenicity of ENIGMA BRCA2 variants with an accuracy of 99.9%. The model also performed excellently on predicting the functional consequence of the independent set of variants (accuracy was up to 91.3%).Conclusion: This new, gene-specific model is an accurate method for interpreting the pathogenicity of variants in the BRCA2 gene. It is a valuable addition for variant classification and can prioritize unreviewed variants for functional analysis or expert review.

BRCA1-specific machine learning model predicts variant pathogenicity with high accuracy

“…We chose the XGBoost (Extreme Gradient Boosting) framework for building the gene-specific model because it implements a highly flexible, optimized, distributed gradient boosting machine learning algorithm ( 35 ). Moreover, when compared with other algorithms, it performed well in predicting the functional impact of BRCA1 and BRCA2 missense variants ( 31 , 34 ). We also have recently applied an XGBoost model to predict the ENIGMA benign/pathogenic binary classification for BRCA2 variants, and it outperformed all previous prediction models ( 34 ).…”

“…Moreover, when compared with other algorithms, it performed well in predicting the functional impact of BRCA1 and BRCA2 missense variants ( 31 , 34 ). We also have recently applied an XGBoost model to predict the ENIGMA benign/pathogenic binary classification for BRCA2 variants, and it outperformed all previous prediction models ( 34 ).…”

“…The other variables that were also collected from VEP included the position of the variant, number of bases involved based on reference and alternative alleles (“variant length”), variant association with phenotype, presence as a somatic mutation, variant impact, and variant consequence. As previously, the variant consequence variables were ranked based on the assumed pathogenicity of the effect with downstream variants having the least effect and stop-gained variants having the highest effect ( 34 ). We obtained the population frequency of the variants from both the BRCA Exchange database and from VEP.…”

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

“…As only a limited number of tools were developed for specific targeted diseases like BC, developing new tools trained on detailed BC data or training existing tools on BC data mostly yields more accurate results for predicting BC pathogenic variants. As proven by several research including Mohannad and Borbala [ 117 ], Nikta et al . [ 76 ] and Hui-Heng et al .…”

“…Finally, Table 9 shows a comparison between the performance of ML-based tools and non-ML-based tools using the AUC values. The gene-specific model tool that is specialized for BC has shown higher AUC compared with other ML-based tools like polyphen-2 and CADD, which, in turn, have shown higher AUC compared with the non-ML based tool SIFT [ 117 ]. Similarly, Lyrus, which is another cancer-specific tool, has shown higher performance in terms of AUC compared with the other ML-based tool Polyphen and the non-ML-based tool SIFT [ 113 ].…”