Abstract Introduction: PARP inhibitors (PARPi) have demonstrated the potential to enhance tumor immunogenicity and sensitize cancer cells to immune checkpoint blockade, as shown in the AMTEC clinical trial. However, identifying PARPi responders and understanding the tumor's adaptive response remain significant clinical challenges. Advances in multi-omic profiling offer unprecedented resolution into these dynamics, yet the complexity and volume of such data demand sophisticated analytical approaches. By applying AI to histopathology images, we aim to uncover predictive morphogenomic biomarkers of PARPi response, potentially enabling more precise treatment selection and to reveal novel mechanisms of therapeutic adaptation in the tumor and its immune microenvironment. Method: We applied two complementary approaches to identify predictive morphometric and morphogenomic biomarkers in biopsy samples collected during the AMTEC trial. Using the PATHOMIQ Phenotype Atlas, we quantified fold changes in individual morphological phenotypes between matched pre-treatment and on-treatment biopsies to identify morphology-based predictors of treatment response. Using the PATHOMIQ Genoscope, a deep learning model that infers RNA expression from H 0.001). Separately, univariate Cox proportional hazards modeling identified 15 genes, predicted directly from H 0.01, log-rank test). Discussion: These results show that AI-based analysis of H 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS5-03-17.
Muhammad et al. (2026) studied this question.