Integrative artificial intelligence and multi-omics modeling approach for characterizing microbial dynamics and health impacts in space microgravity and radiation conditions.

Introduction: Characterizing microbial dynamics and their potential health impacts under space microgravity and radiation conditions remains a major challenge in space biology. The complexity of microbial adaptation, host associated microbiome variation, and heterogeneous multi omics responses requires computational methods that can jointly model temporal dynamics, biological interactions, and predictive uncertainty. Methods: This study introduces an integrative artificial intelligence and multi omics modeling framework, termed the Manifold Aware Event Forecaster, for analyzing microbial behavior and health related outcomes in extreme space environments. The framework consists of three core components: the Counterfactual Dynamics Mapper, the Agent Driven Interaction Planner, and the Uncertainty Weighted Output Filter. different environmental perturbations. The Agent Driven Interaction Planner models microbial community interactions and microbial environment relationships over time. The Uncertainty Weighted Output Filter estimates predictive uncertainty and improves the reliability of health impact prediction. By integrating manifold alignment, interaction modeling, and uncertainty aware aggregation, the proposed framework provides a structured solution for microbial abundance forecasting and health impact assessment under simulated space relevant conditions. Results and discussion: Experimental results show that the proposed approach improves predictive accuracy and interpretability compared with representative machine learning and deep learning baselines. These findings suggest that manifold aware multi omics modeling can support the analysis of microbial adaptation, community dynamics, and health associated risks during long duration space missions.