Deep Learning-Assisted Prioritization of Candidate Drug Targets in Tumors Using Spatial Multi-Omics

Abstract

Simple Summary Cancer treatment is often complicated by the fact that tumor tissues are not uniform. Different cells occupy distinct locations, communicate with nearby cells, and respond differently to treatment. Spatial multi-omics provides a way to examine these differences while preserving information about where biological changes occur in the tissue. Deep learning can further help identify meaningful patterns within these complex datasets and connect molecular changes with specific cell populations and tissue regions. This review discusses how these approaches may support the discovery and prioritization of candidate drug targets, particularly in regions associated with immune suppression, tissue remodeling, invasion, metastasis, and treatment resistance. However, a computationally identified molecule cannot be a validated therapeutic target without further evidence. Laboratory experiments, drug-response studies, organoid and animal models, and clinical investigations are still required. Together, these methods may provide a more efficient and reliable route from spatial observations to experimentally supported therapeutic targets.