Abstract

Introduction: Recently, PARP inhibitors have been approved for treatment of triple negative breast cancer in patients with germline or somatic alterations in BRCA1 or BRCA2. Whilst many mechanisms of action have been proposed, it is likely that the response in this group is due to deficiencies in the homologous recombination repair pathway. Thus, a biomarker that is able to identify patients who may have deficiency in homologous repair despite a lack of mutations in BRCA1 or BRCA2 would have significant clinical utility. Methods: We designed a nearest-centroid classifier for homologous recombination deficiency (HRD) in Asian TNBCs from the Malaysian Breast Cancer (MyBrCa) cohort using an RNA-seq gene expression dataset of 100 genes (HRD100). The classifier was trained on data from 94 TNBC samples from the MyBrCa cohort, and validated in an additional 35 and 87 TNBC samples from MyBrCa and TCGA. The classifier was also validated on the NanoString nCounter platform, as well as using FFPE instead of fresh frozen tissue. Results: The HRD100 classifier identified samples with strong HRD mutational signature at an AUROC of 0.892 in the MyBrCa training dataset, as well as 0.783 and 0.713 in MyBrCa and TCGA validation datasets, respectively. Analysis of the 100 genes in the HRD100 classifier using the NanoString nCounter platform showed a concordance rate of 98% (CI: 95-100%) with RNA-seq gene expression analyses, and a concordance rate of 87% (CI: 73-100%) between FFPE and fresh frozen tissue. Conclusion: Taken together, gene expression using these 100 selected genes may identify triple-negative breast cancer patients with homologous recombination deficiency who may benefit from treatment with PARP inhibitors or platinum chemotherapy.

Keywords: Triple-negative breast cancer, Homologous recombination deficiency, Gene expression signature, PARP inhibitors