A deep-learning pathology model in the British Journal of Cancer aimed to predict nivolumab outcomes in gastric cancer before treatment. Hong, Hwang, Kim and colleagues describe a risk score derived from tumor microscopic features to anticipate response patterns to PD-1 blockade. Pathology-based prediction could help triage patients for immunotherapy and support clinical trial enrichment by identifying likely responders earlier in the care pathway. The study’s emphasis on a pre-treatment scoring framework aligns with precision oncology goals where imaging and genomics are not always immediately available. If validated across cohorts and platforms, the model may support decision-making in clinical settings where turnaround time and reproducibility are key constraints.