Application of long-read sequencing to resolve cryptic genomic instability across cancer types.

3067 Background: Computational mutational signatures can identify tumors with homologous recombination deficiency (HRD), mismatch repair deficiency (MMRD), and focal tandem duplication (FTD) phenotypes, nominating patients for PARP inhibitors or immune checkpoint blockade. However, many signature-positive cases lack identifiable causative alterations by conventional sequencing, forcing patients toward empiric therapy or clinical trial enrollment without molecular confirmation. We evaluated whether long-read nanopore sequencing could provide definitive molecular diagnoses to guide evidence-based treatment selection. Methods: We analyzed 768 patients with advanced prostate (n=505), breast (n=246), and ovarian (n=17) cancers enrolled in the MiOncoSeq precision oncology program. Computational signature analysis identified cases with HRD, MMRD, or FTD phenotypes. Cases unresolved by standard short-read sequencing underwent long-read nanopore sequencing with integrated structural variant detection, methylation profiling, and haplotype phasing. Results: Signature analysis identified 164 cases (21.3%) with genomic instability phenotypes. Short-read sequencing resolved 118/164 cases (71.9%), leaving 46 patients without molecular confirmation. Long-read sequencing achieved mechanistic resolution in 32/46 cases (69.6%), with complete resolution of MMRD (5/5, 100%) and FTD (7/7, 100%). Secondary signature analysis reclassified 11/14 unresolved HRD cases as non-classical, likely reflecting alternative DNA repair defects or replication stress, refining diagnostic yield to 86.9% (20/23) for credible HRD. Cryptic mechanisms included promoter hypermethylation ( BRCA1 , MLH1 , RAD51C ), balanced structural rearrangements ( CDK12 , PMS2 ), and deep intronic splice variants ( PALB2, BARD1 ). Long-read sequencing also uncovered 3 previously undetected pathogenic germline variants ( BARD1 , BRIP1 , PALB2 ) with direct implications for hereditary cancer risk assessment and family screening. Combined analysis identified actionable findings enabling targeted therapy selection in 19.5% of patients. Conclusions: Long-read sequencing provides definitive molecular diagnoses for patients with signature-positive tumors, enables evidence-based therapy selection over empiric treatment, and uncovers cryptic germline variants informing hereditary cancer risk.