Clinical feasibility study of Optim.AI 2.0, a co-culture high-content functional precision platform for predicting combinatorial immunotherapy response

Background: Immunotherapies have advanced cancer treatment, but variable outcomes and the lack of robust predictive biomarkers limit their use. Better patient selection and rational combination strategies could expand their effectiveness. Optim.AI™ is a combinatorial functional precision medicine platform previously shown to identify effective combination treatments for hematologic cancers and sarcoma. Earlier versions guided chemo- and targeted therapy choices but lacked immune components, restricting the drugs assessed. Here, we evaluate the clinical feasibility of Optim.AI™ 2.0, which integrates high-content imaging with tumor-immune ex vivo co-cultures to predict responses to immunotherapy combinations across solid and hematologic tumors. Methods: For Optim.AI™ 2.0, peripheral blood mononuclear cells and tumor cells (non-Hodgkin lymphomas, gynecological cancers or gastrointestinal cancers) were fluorescently labeled to facilitate cell tracking. Combinatorial drug treatment was carried out on co-culture models with indication-specific 12-drug panels, including monoclonal antibodies, antibody-drug conjugates and bispecific antibodies. High-content imaging analysis of tumor-specific cell death was evaluated for Optim.AI™ 2.0 analysis, which searches 531,441 possible permutations derived from 155 ex vivo test combinations to predictively rank all clinically actionable treatments within the drug panel. Clinically relevant immunotherapies were compared with patient-specific factors-disease stage, subtype, and treatment history-to assess concordance with predicted responses. Results: Optimized effector-to-target ratios were established to effectively quantify immune-mediated tumor killing, including antibody-dependent cellular cytotoxicity. High-content imaging captured tumor-specific killing and key immune-tumor interactions such as immune cell migration and tumor infiltration. Across multiple indications, Optim.AI™ 2.0 demonstrated feasibility by detecting antigen-dependent responses and identifying context-specific immunotherapy combinations in both hematologic and solid tumor models. It accurately predicted sensitivity or resistance to first-line rituximab in naïve and relapsed/refractory diffuse large B-cell lymphoma and revealed additional immunotherapy combinations with potential utility in later treatment lines. Conclusion: We developed a high-content functional analytics platform that assesses immunotherapy drug sets in a physiologically relevant tumor-immune co-culture system. With further validation, it could help clinicians select effective immunotherapies. When paired with molecular profiling, Optim.AI™ 2.0 may identify novel immunotherapy combinations for specific patient populations defined by known or emerging biomarkers.