Abstract Introduction: Homologous recombination deficiency (HRD) is a predictive biomarker of response to PARP inhibitors and other DNA damage response (DDR)-targeting therapies across cancer types. HRD can be detected by identifying mutations in BRCA1/2 or other homologous recombination repair (HRR) genes, or through copy number (CN) or single nucleotide variant (SNV)-based signatures associated with genomic instability. While genomic instability status (GIS) is an FDA-approved biomarker for ovarian cancer treatment, comprehensive tissue-based GIS detection across multiple cancers is still limited. We developed and validated a tissue-based GIS predictor integrating CN and SNV signatures for breast, ovarian, and pancreatic cancers. Methods: We implemented an ensemble model to predict GIS using HRR gene deficiency to infer SNV and CN signatures associated with HRD. The model was trained on clinical samples processed on Guardant360 Tissue (Guardant Health, Palo Alto, CA), incorporating select HRD-associated COSMIC single base substitution (SBS) signatures (v3.4) and genomic scarring features such as loss-of-heterozygosity (LOH). The aggregate model was evaluated on an independent cohort of clinical breast, ovarian, and pancreatic samples with tumor purity ≥ 20%. BRCA1/2 or PALB2 biallelic loss and strict wildtype status in 14 HRR genes were used as orthogonal labels to assess sensitivity; while a cohort wildtype for an expanded set of pathogenic or likely pathogenic variants in HRR genes was used to estimate specificity. Results: The GIS model achieved an overall AUC of 0.98, with sensitivity of GIS detection of 90% in breast cancer, 91% in ovarian cancer, and 80% in pancreatic cancer. GIS detection in HRR-negative samples was 3% in all three cancer types. Prevalence of GIS detection among clinical samples was 256/803 (32%) in breast cancer, 56/107 (52%) in ovarian cancer, and 40/214 (19%) in pancreatic cancer. Among patients with GIS detected, 75% of breast cancers, 39% of ovarian cancers, and 67% of pancreatic cancers did not harbor pathogenic BRCA1/2 or PALB2 mutations. Conclusions: We present a tissue-based genomic instability detection method demonstrating strong performance in breast, ovarian, and pancreatic cancers. Prostate cancers, which also commonly exhibit HRD, showed a distinct genomic scarring signature and are being actively investigated for future incorporation into the GIS model.This comprehensive biomarker expands HRD testing beyond HRR mutations to capture a broader patient population who may benefit from DDR-targeting therapies. Citation Format: Pegah Safabakhsh, Denis Tolkunov, Brooke Overstreet, Martina Lefterova, Lauren Lawrence. Tissue-based homologous recombination deficiency status prediction in patients with breast, ovarian, and pancreatic cancers abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 4016.
Safabakhsh et al. (Fri,) studied this question.