Broad-coverage targeted lipidomics reveals stage-specific lipid metabolic profiles and diagnostic and prognostic marker panels in ischemic stroke
Haoran Huang, Liu Liu, Can Zhao, Shuying Cheng, Wenxuan Li, Anbao Xu, Xiaoqin Yin, Xin Xu
BACKGROUND: Ischemic stroke (IS) is a major cause of mortality and disability globally, with challenges in early diagnosis and prognosis prediction. Dysregulated lipid metabolism is key to IS pathophysiology, but comprehensive profiling of lipid changes during disease progression remains limited. METHODS: This study enrolled 223 IS patients and 57 healthy controls. Plasma lipid profiles were analyzed using broad-coverage targeted lipidomics by ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS). Differential lipids were identified through orthogonal partial least squares discriminant analysis (OPLS-DA) with univariate analysis, and their changes across acute, subacute, convalescent, and chronic phases were examined by clustering analysis. Machine learning algorithms, including least absolute shrinkage and selection operator (LASSO) regression and support vector machine (SVM), were used to screen diagnostic and prognostic biomarkers, followed by logistic regression models and receiver operating characteristic (ROC) curve evaluation. RESULTS: From 607 identified lipids, 54 showed differential abundance between IS patients and healthy controls, grouped into four clusters. For diagnosis, six lipids-LPG(18:0), PE(O-16:0/18:2), TG(52:2/FA16:0), PE(O-16:0/22:6), PE(O-16:0/20:3), and PE(O-18:0/18:2)- achieved an area under the curve (AUC) of 0.984. For prognosis, six lipids-PE(O-16:0/22:5), PE(O-18:0/22:5), SM(d18:1/14:0), PG(18:0/18:1), PE(O-16:0/20:3), and LPI(16:0)-achieved an AUC of 0.925. CONCLUSION: This study characterizes lipid metabolism changes across IS stages reconstructed from cross-sectional data of different patient groups and establishes two six-lipid panels for diagnosis and prognosis. These findings provide insights into lipid metabolism evolution following stroke and offer candidate biomarker panels for IS management.