HRDPath: an explainable multi-model deep learning architecture for predicting homologous recombination deficiency from histopathology images.

HRDPath predicts homologous recombination deficiency in high-grade serous ovarian carcinoma with an AUC of 0.846, outperforming previous methods across three patient cohorts.

• Why it matters: Reliable biomarkers for HRD are essential for guiding treatment in ovarian cancer, but current genomic tests are costly, variable, and slow, limiting clinical utility.
• What they did: The study developed HRDPath, a deep learning model that analyzes whole slide images from patients, incorporating multiple slides and optimization strategies to improve prediction accuracy.
• The result: HRDPath achieves both sensitivity and specificity of 0.90, providing an interpretable, scalable digital biomarker that enhances treatment stratification and offers biological insights into HRD-positive cancers.