Decoding astrocytic tryptophan metabolism in the pathogenesis of epilepsy: evidence from artificial intelligence-driven multi-omics and clinical validation.

Background: Epilepsy (EP) is a prevalent neurological disorder with complex etiology, often involving metabolic dysregulation. Emerging evidence highlights the role of tryptophan metabolism (TM) and astrocyte dysfunction in EP pathogenesis. This study aimed to decode the TM and astrocyte (TA)-related molecular signature and identify a central therapeutic target for EP. Methods: By integrating seven hippocampal bulk profiles (GSE28674, GSE256068, GSE57585, GSE163296, GSE63808, GSE90886, and GSE134697) from patients with EP using integrative bioinformatics pipelines, including limma, xCell, WGCNA, machine learning, and consensus clustering, we identified a TA-associated diagnostic signature and a molecular stratification model for EP. Next, the TA-associated hub gene was identified by SHAP analysis, and its molecular patterns in astrocytes were characterized using single-cell hippocampal data from patients with EP (GSE190452). In addition, the active-learning framework DrugReflector and molecular docking were used to identify a potential therapeutic agent targeting the TA-associated hub gene in the integrated GSE63808 and GSE90886 dataset. Finally, hippocampal tissue from patients with EP was used to examine expression of the TA-associated hub gene. Results: Four TA-associated shared DEGs were identified: CAT, TXNDC2, RBM27, and GPT. These four DEGs showed predictive value for EP. RBM27 was identified as the central pathogenic factor; it was upregulated and predominantly expressed in astrocytes. Drug prediction identified BRD-K04111260 as a promising compound targeting RBM27 for the treatment of EP. Conclusion: This study establishes a novel TA-associated molecular axis in EP pathogenesis, identifying RBM27 as a critical hub gene with strong diagnostic potential and therapeutic promise.