Comprehensive comparison of homologous recombination deficiency predictors in early-stage triple-negative breast cancer

Systematic comparison of seven HRD prediction methods (DNA, RNA, functional RAD51, image-based) on an early-stage triple-negative breast cancer multi-omics cohort, benchmarked against an FDA-approved assay

This study systematically compares seven homologous recombination deficiency (HRD) prediction methods on a cohort of 235 patients with early-stage triple-negative breast cancer, profiled by RNA-seq, WGS, RAD51 foci, and digital histology images. Sequencing-based methods (HRDetect, CHORD) and scarHRD show the best agreement, while RNA- and image-based discordances appear linked to molecular subtype and training-cohort context. Despite classification variability, all seven methods show comparable prognostic performance in patients on adjuvant chemotherapy. The authors stress the need to rigorously optimize data preprocessing and thresholds to ensure comparability and reproducibility.

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

HRD status governs access to PARP inhibitors, and discordance between predictors is a concrete problem for laboratories that must choose a method. The finding that preprocessing and thresholds explain a large share of the discordance is a strong practical message, more useful than yet another tool. The scope remains somatic (tumor status) and does not directly touch germline predisposition, making this a bioinformatics benchmark rather than an oncogenetics paper.