AI-Enabled Prediction of Homologous Recombination Deficiency from Histopathology Images

Identification of patients who will respond to Poly (ADP-ribose) polymerase (PARP) in-hibitor therapy is challenging due to the lack of a unifying morphological pheno-type. Homologous recombination deficiency (HRD), a key biomarker in breast and ovarian cancer, guides therapeutic decisions. Current HRD testing via next-generation sequencing (NGS) is tissue-dependent, has high failure rates, misses relevant HRD genes, and in-volves longer turn-around times. To overcome these limitations, we devel-oped OncoPredikt, a deep learning model trained on Hematoxylin and Eosin (H&E)-stained whole slide images (WSIs) to non-invasively predict HRD status. Based on a ResNet-50 architecture, the model was trained and validated on 514 WSIs, including 315 breast and 80 ovarian cancer samples from The Cancer Genome Atlas (TCGA), with HRD labels derived from genomic data. An independent dataset of 119 ovarian cancer cases with known BRCA1/2 or HRR mutations was used for validation. OncoPredikt achieved an AUC of 0.85 in breast cancer and 0.97 in ovarian cancer, with robust sensitivity, speci-ficity, and F1-scores in identifying HRD-positive cases. These findings demonstrate On-coPredikt's potential as a rapid, cost-effective, and tissue-sparing alternative to conven-tional NGS testing. While promising, further validation is needed to establish its general-izability across broader cancer types.