Correlation between gene expression and mutator phenotype predicts homologous recombination deficiency and outcome in ovarian cancer.

unlabelledNew strategies are needed to predict response to platinum-based chemotherapy and outcome of ovarian cancers. We hypothesized that the mutator phenotype in the cancer genome represents the overuse of alternative DNA repair mechanisms, which might be a sign of homologous recombination (HR) deficiency and can be captured by gene expression. Multidimensional data of ovarian cancer patients and breast cancer patients from The Cancer Genome Atlas (TCGA) database were used for the development and validation of a potential clinical information-independent score that correlates with HR deficiency and predicts outcome. Correlation of the score with platinum response, outcome, and BRCA mutations was assessed. The score correlated with increased genomic mutation rate in both ovarian cancer and breast cancer cases that harbored a substantial subset of HR-deficient samples. Significantly improved outcomes were observed in the high-scoring group versus the low-scoring group in the TCGA dataset and in three large gene expression microarray datasets. A strong correlation was found between the score and the likelihood of achieving complete response to chemotherapy. The score was also found to be highly robust to noises in genomic mutations. Sixty-four patients harboring BRCA mutations were successfully divided into two groups based on scores, with the high-scoring group showing significantly improved outcomes compared with wild-type cases and the low-scoring group showing no significance in all the same analyses. The score was significantly correlated with the response to platinum therapy and outcome. Evaluation of the score as a prognostic tool in ovarian cancer patients is warranted. KEY MESSAGE: We develop a diagnostic signature for the HR-deficiency based on a novel hypothesis. HR-deficiency score is significantly correlated to platinum therapy and outcomes. HRDS was validated by its association with OS, PFS, DFS and CR in validation datasets. Evaluation of the score as a prognostic tool in ovarian cancer patients is warranted.