An interdisciplinary team led by Ge, Mo and Wei published a Nature Communications study introducing a biologically informed deep‑learning framework that improves prediction of drug response in breast cancer models. The model integrates drug molecular features with pathway‑aware tumor representations to enhance interpretability and forecasting of therapeutic sensitivity across datasets. Authors validated the approach on cell lines and patient cohorts, demonstrating improved performance over baseline models and identifying candidate drug‑pathway interactions for follow‑up. The study positions AI‑driven integrative modeling as an accelerating tool for preclinical prioritization and biomarker discovery in oncology.