DNA repair phenotypes inform carboplatin and PARP inhibitor response

High-grade ovarian carcinoma (HGOC) is characterized by multilayered genomic complexity driven by early disruption of the p53 pathway which enables diverse mutational processes and chromosomal instability (CIN). The molecular consequences of defective DNA repair and replication have been used to derive mutational signatures to predict mutational processes operative in each HGOC. Clinical testing for homologous recombination deficiency (HRD) signatures does not reliably guide the use of platinum-based therapy or poly(ADP-ribose) polymerase inhibitors (PARPi) for women with HGOC. Previous studies have shown that multiple HRD signatures are operative at the single-base substitution, indel, and structural variant levels in the same tumors. However, separate signatures do not capture the continuum of erroneous repair programs active in BRCA1 - and BRCA2 -deficient tumors. Here, we propose a new integrative mathematical approach to recover all naturally occurring genomic features associated with operational DNA repair pathways. We established a national healthcare outcome study of 466 HGOC patients, integrating tumor-normal deep whole-genome sequencing (WGS) with real-world treatment and outcomes. We demonstrated that seven integrative genomic features (IGFs) capture the mutational programs operative in HGOC tumors. We validated the clinical relevance of these IGFs using causal-inference methods. IGF2 captured microhomology‑flanked deletions, templated insertions (TINs), and interstitial deletions. IGF5 captured 1 kilobase (kb) to multi‑megabase (Mb) tandem duplications (TDs) and blunt deletions. IGF8 was rich in inversions and multi‑Mb TDs, whereas IGF10 linked CpG transversions with TINs. IGF2 reflects polymerase theta-mediated end joining, associated with BRCA2 inactivation, favorable prognosis after first-line platinum, and reduced benefit from maintenance PARP inhibitors. IGF5, marked by tandem duplications and non-homologous end joining, is enriched in BRCA1 -inactivated tumors and predicts earlier relapse but increased benefit from maintenance PARPi. In Cox models (per 1 standard deviation unit), IGF2 improved overall survival (OS) (HR 0.81, 95% CI 0.71-0.92) and IGF8 was adverse (HR 1.25, 1.08-1.46), and IGF10 predicted benefit at second‑line relapse when half of patients received maintenance PARPi (progression-free survival HR 0.75, 0.62-0.89). IGF2 modified second-line PARPi benefit ( IGF2-binary x PAPRi interaction P=0.042), with benefit confined to IGF2‑low. This WGS-based taxonomy reframes HRD as a continuum of DNA double-strand break repair choices. The mechanism-specific, WGS-derived phenotypes can be replicated in clinical practice, providing a basis for biomarker-stratified trials of PARP inhibitors and emerging polymerase theta (POLQ) inhibitors in ovarian cancer and in other tumors with DNA-repair defects.