Machine learning-based immune cell predictive model for cytokine-induced killer cell therapeutic efficacy in gastrointestinal cancer* 3340

Abstract Description Cancer remains the leading cause of death worldwide, driving a substantial global demand for effective cancer cell therapies. Despite advancements, physicians face challenges in identifying the most suitable treatment for individual patients due to the lack of standardized methodologies. In this study, peripheral blood mononuclear cells from cancer patients were collected prior to cytokine-induced killer (CIK) immunotherapy and analyzed using flow cytometry with a multi-antibody panel to identify immune cell populations. Post-therapy, patient responses were evaluated based on the Response Evaluation Criteria in Solid Tumors (RECIST) to assess therapeutic efficacy. The objective was to develop machine learning prototypes—such as principal component analysis (PCA), support vector machines (SVM), K-nearest neighbor (KNN), and Decision Tree (DT) models—to classify immune cell data into distinct RECIST groups and predict therapeutic outcomes. Among these, PCA and DT models demonstrated accuracy rates exceeding 85%. Integrating immune monitoring with AI technologies has the potential to enhance preventive medicine and therapeutic diagnostics, paving the way for advancements in precision medicine. Keywords Immune monitoring, CIK, AI, Machine learning, Precision medicine Funding Sources *This work is supported by a grant from the National Science and Technology Council (113-2314-B-038-137), Taipei, Taiwan. Topic Categories Tumor Immunology: Checkpoints, Prevention, and Treatment (TIPT)