Predicting ovarian/breast cancer pathogenic risks of BRCA1 gene variants of unknown significance

The difficulty of early diagnosis for ovarian cancer is an important cause of the high mortal rates of ovarian cancer patients. Instead of symptom-based diagnostic methods, modern sequencing technologies enable the access of humans genetic information via reading DNA/RNA molecules nucleotide base sequences. In such way, genes mutations and variants could be identified and hence a better clinical diagnosis in molecular level could be expected. However, as sequencing technologies gain more popularity, novel gene variants with unknown clinical significance are found, giving difficulties to interpretations of patients genetic data, precise disease diagnoses as well as the making of therapeutic strategies and decisions. In order to solve these issues, it is of critical importance to figure out ways to analyze and interpret such variants. In this work, BRCA1 gene variants with unknown clinical significance were identified from clinical sequencing data, and then we developed machine learning models so as to predict the pathogenicity for variants with unknown clinical significance. Amongst, in performance benchmarking, our optimized random forest model scored 0.85 in area under receiver-operating characteristic curve, which outperformed other models. Finally, we applied the optimized random forest model to predict the pathogenic risks of 7 BRCA1 variants of unknown clinical significances identified from our sequencing data, and 6315 variants of unknown clinical significance in ClinVar database. As a result, our model predicted 4724 benign and 1591 pathogenic variants, which helped the interpretation of these variants of unknown significance and diagnosis.