A machine learning model integrating gut microbiota biomarkers for predicting neurological recovery after cerebral hypoxia-ischemia: a single-center study augmented with public data

Objective: To evaluate the predictive value of gut microbiota biomarkers for neurological recovery after cerebral hypoxia-ischemia and to develop a clinically oriented, validated machine learning (ML) for early outcome prediction. Methods: In this single-center cohort study, 772 patients (2022-2024) were stratified according to 1-year neurological outcomes into favorable (n = 538) and poor recovery (n = 234) groups, gut microbiota composition, microbial metabolites, systemic inflammatory and oxidative stress markers, and selected gene expression profiles were assessed using blood and fecal samples. Microbial profiling was assessed via 16S rRNA sequencing and whole-genome sequencing. LASSO regression, random forest (RF), and neural networks models were constructed and evaluated using internal validation strategies and augmented with publicly available datasets. Results: ). Across ML approaches, IL-6 consistently emerged as a key predictive feature, with RF confirmed its prelative importance. The neural network model showed stable predictive performance across validation analyses, indicating robustness for high-dimensional biomarker integration. Conclusion: Gut microbiota-associated biomarkers, when combined with systemic inflammatory indicators such as IL-6, enable robust early prediction of neurological recovery following cerebral hypoxia-ischemia. ML-based integration of multi-modal biomarkers may facilitate clinically applicable prognostic assessment, while also providing a foundation for future investigations into microbiota-targeted interventions.