Immunotherapy drug target identification using machine learning and patient-derived tumour explant validation

Immunotherapy has revolutionised cancer treatment, yet few patients respond clinically, necessitating alternative strategies that can benefit these patients. Novel immune-oncology targets can achieve this through bypassing resistance mechanisms to standard therapies. To address this, we introduce MIDAS, a multimodal graph neural network system for immune-oncology target discovery that leverages gene interactions, multi-omic patient profiles, immune cell biology, antigen processing, disease associations, and phenotypic consequences of genetic perturbations. MIDAS generalises to time-sliced data, outcompetes existing methods, including OpenTargets, and distinguishes approved from prospective targets. Moreover, MIDAS recovers immunotherapy response-associated genes in unseen trials, thus capturing tumour-immune dynamics within human tumours. Interpretability analyses reveal a reliance on autoimmunity, regulatory networks, and relevant biological pathways. Functionally perturbing the OSM-OSMR axis, a proposed target, in TRACERx melanoma patient-derived explants yielded reduced dysfunctional CD8 ⁺ T cells, which associate with immunotherapy response. Our results present a machine learning framework for analysing multimodal data for immune-oncology discovery.