Hybrid Mechanistic–AI Modeling with Physics-Trained Neural Networks for Patient Digital Twins

Patient digital twins aim to create computational replicas of an individual’s physiology that can predict disease trajectories and treatment response. Today, many digital twin approaches sit at one of two extremes: mechanistic models (built from biology and physics, interpretable but often hard to personalize and computationally heavy) or purely data-driven AI (flexible and accurate in-distribution, but often brittle, data-hungry, and harder to trust clinically). This Research Topic focuses on hybrid mechanistic–AI modeling, where physics trained neural networks (e.g., physics informed neural networks and related constrained architectures) are used to integrate first principles knowledge with patient data to build robust, personalized, and clinically meaningful digital twins.We invite contributions that develop, evaluate, or translate hybrid frameworks combining differential equation models, multiscale physiology, and machine learning methods that respect physical laws, biological constraints, and clinical priors. Emphasis is placed on approaches that improve generalization under limited data, enable uncertainty quantification, support causal reasoning, and maintain interpretability and safety—all essential for real-world clinical decision support.Key themes include, but are not limited to: (1) neural surrogates for mechanistic simulators; (2) physics- or physiology-constrained learning for time-series and spatiotemporal patient data; (3) data assimilation, parameter inference, and personalization of mechanistic models using deep learning; (4) multimodal digital twins integrating imaging, wearables, -omics, and EHR data; (5) uncertainty-aware prediction, validation, and benchmarking; and (6) deployment challenges such as model governance, privacy, fairness, regulatory considerations, and clinical workflow integration.This Research Topic aims to advance the methodological foundations and translational readiness of patient digital twins by uniting mechanistic insight with AI scalability, accelerating the move from proof-of-concept models to trustworthy, patient-specific predictive tools for precision medicine. Topic editors: Mohammadreza Nikmaneshi, Narsis Aftab Kiani Source: https://www.frontiersin.org/research-topics/78702/hybrid-mechanistic-ai-modeling-with-physics-trained-neural-networks-for-patient-digital-twins