Abstract
Inappropriate perioperative anesthesia (PA) drug management may lead to the detrimental clinical outcomes of Kawasaki disease (KD) patients with deep vein thrombosis (DVT). However, perioperative anesthesia-related drug target genes(PADTGs) role in KD patients with DVT has not yet been elucidated. This study aims to decipher PARDTGs molecular patterns in KD patients with DVT via machine learning-assisted multi-omics. By integrating 3 DVT bulk datasets (GSE118259, GSE19151, GSE48000) and 3 machine learning pipelines(RF, Lasso and SVM-RFE), we discovered the PARDTGs-associated DEGs and hub genes for DVT patients. Consensus clustering and nomogram were utilized for identifying PARTGs-associated molecular subgroups for DVT patients in GSE19151 and dignostic model for DVT and KD patients in aforementioned 3 DVT bulk datasets and 2 KD bulk profiles(GSE68004 and GSE18606). Indeed, hub gene molecular and immune features were estimated in DVT bulk profiles(GSE19151) and KD single-cell data(GSE254657) at temporal and spatial manners. Besides, Therapeutic agent targeting hub gene in purpose of reversing DVT and KD to healthy status were enriched by CTD database, and validated by molecular docking. In addition, an in vitro KD model further validated hub gene expression. PTMA and FTHL8 were identified as 2 PARDTGs-associated signatures involved in KD patients for forecasting DVT pathogenesis, which was mainly distributed in neutrophils of KD patient peripheral blood. Besides, Gallic acid can be considered as reproposing approaches for treatment DVT in KD patients. Our study systemically discovered PARDTGs-related signature involved in DVT pathogenesis of KD patients via machine learning-driven multi-omics, which provides novel idea in their clinical application. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s10616-026-00946-4.