2025/05/10 by Zhixuan Wang, Wang, Zhixuan
Biochemistry, Genetics and Molecular Biology · Medicine · #FOS: Biological sciences #FOS: Computer and information sciences #Genetics, Bioinformatics, and Biomedical Research #Machine Learning (cs.LG) #Machine Learning in Bioinformatics #Neurons and Cognition (q-bio.NC) #Traditional Chinese Medicine Studies
paper · pdf · doi:10.48550/arxiv.2505.09643
openalex publication_date 2025/05/10 · openalex created_date 2025/10/15 · openalex updated_date 2026/07/28
Epilepsy is a prevalent neurological disease with millions of patients worldwide. Many patients have turned to alternative medicine due to the limited efficacy and side effects of conventional antiepileptic drugs. In this study, we developed a computational approach to optimize herbal epilepsy treatment through AI-driven analysis of global natural products and statistically validated randomized controlled trials (RCTs). Our intelligent prescription system combines machine learning (ML) algorithms for herb-efficacy characterization, Bayesian optimization for personalized dosing, and meta-analysis of RCTs for evidence-based recommendations. The system analyzed 1,872 natural compounds from traditional Chinese medicine (TCM), Ayurveda, and ethnopharmacological databases, integrating their bioactive properties with clinical outcomes from 48 RCTs covering 48 epilepsy conditions (n=5,216). Using LASSO regression and SHAP value analysis, we identified 17 high-efficacy herbs (e.g., Gastrodia elata [using é for accented characters], Withania somnifera), showing significant seizure reduction (p<0.01, Cohen's d=0.89) with statistical significance confirmed by multiple testing (p<0.001). A randomized double-blind validation trial (n=120) demonstrated 28.5% greater seizure frequency reduction with AI-optimized herbal prescriptions compared to conventional protocols (95% CI: 18.7-37.3%, p=0.003).