Pan-Cancer Analysis of Homologous Recombination Deficiency in Cell Lines

Homologous recombination deficiency (HRD) drives genomic instability in multiple cancer types and renders tumors vulnerable to certain DNA-damaging agents such as PARP inhibitors. Thus, HRD is emerging as an attractive biomarker in oncology. A variety of in silico methods are available for predicting HRD; however, few of these methods have been applied to cell lines in a comprehensive manner. Here, we utilized two of these methods, “Classifier of HOmologous Recombination Deficiency” and “HRDsum” scores, to predict HRD for 1,332 cancer cell lines and 84 noncancerous cell lines. Cell lines with biallelic mutations in BRCA1 or BRCA2, which encode key components of the homologous recombination pathway, showed the strongest HRD predictions, validating the two methods in cell lines. A small subset of BRCA1/2 wild-type cell lines was also classified as HRD, several of which showed evidence of epigenetic BRCA1 silencing. Similar to HRD in patient samples, HRD in cell lines was associated with p53 loss, was mutually exclusive with microsatellite instability, and occurred most frequently in breast and ovarian cancer types. In addition to validating previously identified associations with HRD, we leveraged cell line–specific datasets to gain new insights into HRD and its relation to various genetic dependency and drug sensitivity profiles. We found that in cell lines, HRD was associated with sensitivity to PARP inhibition in breast cancer but not at a pan-cancer level. By generating large-scale, pan-cancer datasets on HRD predictions in cell lines, we aim to facilitate efforts to improve our understanding of HRD and its utility as a biomarker.

HRD is common in cancer and can be exploited therapeutically, as it sensitizes cells to DNA-damaging agents. Here, we scored more than 1,300 cancer cell lines for HRD using two different bioinformatic approaches, thereby enabling large-scale analyses that provide insights into the etiology and features of HRD.