Artificial Intelligence for Mapping Cellular and Neural Circuit State Transitions in Human Disease: Toward Multiscale Disease Modeling

Understanding human disease requires frameworks that can connect molecular and cellular perturbations to systems-level dysfunction and clinical phenotype across time. Recent advances in single-cell and spatial profiling have revealed that cellular identity is dynamic and distributed across continuous state landscapes, while developments in electrophysiology, neuroimaging, and circuit analysis have underscored the importance of distributed neural dynamics in health and disease. In parallel, artificial intelligence (AI) has emerged as a powerful approach for analyzing high-dimensional, multimodal biomedical data and reconstructing biological relationships that are difficult to resolve using conventional methods alone. In this narrative review, we examine how AI-based methods can be used to map cellular and neural circuit state transitions in human disease and how these approaches may be integrated to support multiscale models of disease progression. We discuss current strategies for characterizing cellular state landscapes, including single-cell, spatial, trajectory-based, graph-based, and multimodal approaches, and we review AI-driven methods for decoding neural circuit dynamics from electrophysiological, imaging, and large-scale functional datasets. We further highlight bidirectional interactions linking cellular states, synaptic and microcircuit remodeling, circuit-level dysfunction, and behavioral or clinical outcomes, emphasizing disease progression as a sequence of coupled transitions across molecular, cellular, synaptic, and network scales. Finally, we discuss the implications of AI-integrated multiscale medicine for biomarker discovery, disease trajectory modeling, therapeutic window identification, and adaptive precision intervention, while addressing challenges related to causality, interpretability, validation, ethics, and clinical translation. Together, these developments support a shift from static classifications of disease toward dynamic, multiscale, and clinically relevant models that better reflect the evolving behavior of biological systems.