A review in the Journal of Translational Medicine argues that AI-derived biomarkers could improve prediction of cancer immunotherapy response beyond conventional single-marker tests such as PD-L1, tumor mutational burden, and microsatellite instability. The authors highlight that deep learning approaches in digital pathology and radiomics can quantify tumor-infiltrating lymphocytes and spatial immune phenotypes, while multi-omics models (including graph neural networks and transformers) attempt to capture tumor-immune interaction complexity that traditional biomarker panels miss. The review also notes caution: reported discriminatory performance varies by cancer type, endpoints, and validation design, and explainability outputs may not be causal. Still, the paper positions AI biomarkers as increasingly relevant for stratifying immunotherapy candidates and anticipating immune-related adverse events.