Cancer’s genetic heterogeneity complicates treatment due to varied drug responses and resistance. Combination therapy improves efficacy but is challenging due to the vast number of potential drug combinations. Machine learning offers promise in predicting drug synergy and optimizing treatment; however, current models struggle with high-dimensional data, multi-omics integration, and accurately capturing drug synergy based on multiple synergy metrics and drug dose–response data. This review discusses advancements in AI-driven drug combination optimization, emphasizing the necessity of incorporating multi-omics data, multiple synergy metrics, dosage integration, and the need to incorporate patient-specific factors, treatment histories, and resistance mechanisms to refine chemotherapy regimens, ultimately facilitating multi-omics data integration for personalized cancer therapies.