A deep-learning pathology model reported in the British Journal of Cancer aims to predict nivolumab outcomes in gastric cancer using tumor microscopic features. The study describes a risk-score approach designed to forecast response before immunotherapy begins. Pathology-based predictive analytics are increasingly being tested as a way to match therapies to tumor biology, potentially improving outcomes and reducing ineffective treatment exposure. Here, the focus remains on pre-treatment stratification for PD-1 blockade. If validated prospectively, such models could become decision-support tools integrated into pathology workflows—bridging molecular phenotyping and clinical treatment selection.