A Multi-Scale Integrative Framework Identifies THBS1 as a Master Diagnostic Hub and Prioritises Small-Molecule Therapeutics for Subarachnoid Hemorrhage

Background Subarachnoid hemorrhage resulting from intracranial aneurysm rupture is a devastating cerebrovascular event with an exceptionally grim prognosis. Identifying the cross-scale molecular mechanisms driving quiescent aneurysmal wall degeneration may facilitate the early diagnosis of high-risk lesions and contribute to targeted medical therapy development. Methods Three transcriptomic datasets (GSE13353, GSE15629, and GSE54083) were obtained from the GEO database. After batch effect correction using ComBat, the GSE13353 and GSE15629 datasets were integrated for analysis, and differentially expressed genes were identified using the limma and WGCNA frameworks. Subsequently, three machine learning feature selection algorithms, namely LASSO, SVM-RFE, and Random Forest, were applied to the external validation dataset (GSE54083) to identify a core set of diagnostic genes reproducible across cohorts and to construct clinical prediction models. Single-cell transcriptomic profiles were utilized to investigate the cellular origins of these core genes and to elucidate the landscape of myeloid cell heterogeneity within the disease microenvironment. Furthermore, we integrated a cutting-edge AI-driven framework (DrugReflector) with blind molecular docking, molecular dynamics simulations, MM-GBSA-based free energy calculations, and density functional theory to identify optimal potential therapeutic candidates based on integrated disease expression profiles. Results Thrombospondin 1 (THBS1) has been identified as the sole cross-cohort, reproducible diagnostic core gene to which all machine learning models converged. Furthermore, dihydrosamidin and BRD-K22096725 can be considered as complementary potential therapeutic agents for the treatment of intracranial aneurysms by stably targeting THBS1 through distinct hydrophobic and electrostatic mechanisms. Conclusion By integrating AI pipelines and multi-omics, our study first traced THBS1 mechanisms in intracranial aneurysm pathogenesis and elaborated on THBS1's predictive and druggable potential for patients at risk of subarachnoid hemorrhage.