Identification of methylation signatures to assess homologous recombination deficiency in breast cancer

DNA methylation is the most studied epigenetic modification in cancers. Methylation signatures from tissue samples and liquid biopsies can help in the identification of cancer subtypes and are used for prognostication and treatment decision. Homologous recombination deficiency (HRD) is a characteristic of some cancers related to inefficient DNA damage repair and is measured through the analysis of mutations in specific genes and global genomic aberrations. Breast cancers with HRD are more sensitive to certain treatments, therefore HRD evaluation is an important aid in treatment selection. In a publicly available dataset of breast cancer samples with matched genomic and methylation data (TCGA-BRCA), we identified sites with differential methylation between HRD-high and HRD-low samples. Using bioinformatic tools, we found methylation signatures that could identify HRD cancers, and we created different models (gaussian and binomial) that could classify breast cancer samples into HRD classes. We then validated the models in another independent dataset of breast cancers (SCAN-B), demonstrating the feasibility to use methylation signatures to infer HRD status. Given the feasibility of methylation tests in liquid biopsies even with low amount of cancer free DNA, these models could help in identifying breast cancer patients that would benefit from HRD-directed therapies using blood-based test, thus enhancing the ability to recognize those patients in the clinical setting.