Context-dependent immune regulation: a mechanistic and AI-enabled integrative framework

The immune system functions as a dynamic, multiscale, and context-sensitive network whose outputs depend on cellular composition, spatial organization, metabolic state, host background, and temporal trajectory. These properties have long been investigated through systems immunology, mathematical modeling, and immune simulation. This Perspective uses context dependence as an organizing principle to connect these foundations with recent artificial intelligence (AI) methods, and develops a context-centered view in which immune context is treated as an explicit, partially measurable, and partially uncertain state that can be encoded, constrained, and prospectively tested. It summarizes representative phenomena across cancer, infection, innate immunity, autoimmunity, and genetic variation; describes their microenvironmental, metabolic, systemic, and temporal determinants; and distinguishes the capabilities of mechanistic, data-driven, symbolic, and hybrid models. The proposed hybrid AI-mechanistic framework positions AI as a tool for context representation and multimodal integration, while mechanistic and knowledge-based components preserve explicit dynamics, biological constraints, and testable interventions. Practical strategies are discussed for limited biomaterial, partially longitudinal animal studies, missing or unpaired modalities, interpretability, and prospective validation. The framework does not require every modality in every sample; instead, it emphasizes minimum context annotation, uncertainty-aware integration, explicit assumptions, and benchmarking against simpler models.