AI-Driven Systems Oncology: Modeling Tumor Evolution Through Multimodal Data Integration

Abstract

Cancer remains one of the leading causes of morbidity and mortality worldwide despite remarkable advances in molecular diagnostics, targeted therapeutics, immunotherapy, and precision medicine. Tumor development is a dynamic systems-level process driven by genomic instability, epigenetic alterations, transcriptional regulation, metabolic reprogramming, immune interactions, and continuous adaptation within the tumor microenvironment. Understanding these complex biological processes requires computational approaches capable of integrating heterogeneous biomedical information across multiple molecular and clinical scales. Artificial intelligence (AI)- driven systems oncology has emerged as a transformative discipline that combines systems biology, multimodal learning, foundation AI models, graph neural networks, digital pathology, radiology, multi-omics, longitudinal clinical data, and mathematical modeling to characterize tumor evolution and support precision medicine. Machine learning, deep learning, transformer architectures, self-supervised learning, digital twins, agentic AI, and generative artificial intelligence now enable integration of radiological imaging, genomic sequencing, transcriptomics, proteomics, metabolomics, laboratory biomarkers, wearable technologies, and electronic health records into unified patient-specific computational representations. These intelligent systems support diagnosis, prognostic prediction, biomarker discovery, therapeutic optimization, immunotherapy selection, adaptive treatment planning, and clinical decision support. Despite remarkable progress, important challenges remain regarding multimodal data harmonization, computational scalability, explainability, interoperability, regulatory validation, and ethical governance. This review provides a comprehensive overview of AI-driven systems oncology, highlighting computational principles, clinical applications, implementation challenges, and future perspectives for modeling tumor evolution through multimodal data integration.[1]