AI-driven multi-omics integration in breast cancer: clinical applications, immunotherapy prediction, and translational challenges

Abstract

Breast cancer remains a major global threat to women’s health, and its marked molecular, spatial, and temporal heterogeneity continues to limit the accuracy of early diagnosis, risk stratification, and treatment-response prediction. Conventional single-dimensional diagnostic and therapeutic models are often insufficient to capture the dynamic complexity of the tumor microenvironment (TME). In this context, artificial intelligence (AI) provides powerful tools for high-dimensional feature extraction, multimodal data integration, and cross-scale modeling of heterogeneous biological and clinical information. AI-enabled approaches can link microscopic molecular alterations with macroscopic imaging and pathological phenotypes, while also characterizing complex cellular interactions within the TME. This review summarizes current applications of AI-driven multimodal and multi-omics integration in breast cancer, focusing on early detection, precision diagnosis, immunotherapy-response prediction, drug-sensitivity assessment, and prognostic evaluation. We also discuss key translational challenges, including data heterogeneity, batch effects, interpretability, privacy protection, regulatory considerations, and clinical workflow integration. Overall, AI-driven multi-omics strategies offer a promising framework for improving individualized treatment selection and real-time monitoring in breast cancer, although robust prospective validation and multidisciplinary implementation are still required before routine clinical adoption. Compared with previous modality-specific reviews, this article emphasizes cross-modal evidence appraisal, clinical-maturity stratification, data-quality constraints, foundation model opportunities, and regulatory requirements for translating AI-enabled multi-omics from exploratory research to clinically auditable decision support.