LAUSR

Background The diagnosis of cerebral thrombosis origin is challenging and remains unclear. This study aims to identify thrombosis due to cardioembolism (CE) and large artery atherosclerosis (LAA) from a new perspective of distinct metabolites. Methods Distinct metabolites between 26 CE and 22 LAA origin thrombi, which were extracted after successful mechanical thrombectomy in patients with acute ischemic stroke in the anterior circulation, were analyzed with a ultra performance liquid chromatography coupled with quadrupole time-of-flight mass spectrometry (UPLC-QTOF-MS) system. Enriched metabolic pathways related to the metabolites were identified. Least absolute shrinkage selection operator regression analyses and a filtering method were used to select potential predictors. Furthermore, four machine learning classifiers, including decision tree, logistic regression, random forest (RF), and k means unsupervised classification model, were used to evaluate the predictive ability of the selected metabolites. Results UPLC-QTOF-MS analysis revealed that levels of 88 and 55 metabolites were elevated in LAA and CE thrombi, respectively. Kyoto Encyclopedia of Genes and Genomes analysis revealed a significant difference between the pathways enriched in the two types of thrombi. Six metabolites (diglyceride (DG, 18:3/24:0), DG (22:0/24:0), phytosphingosine, galabiosylceramide (18:1/24:1), triglyceride (15:0/16:1/o–18:0), and glucosylceramide (18:1/24:0)) were finally selected to build a predictive model. The predictive RF model was confirmed to be the best, with a satisfactory stability and prediction capacity (area under the curve=0.889). Conclusions Six metabolites as potential predictors for distinguishing between cerebral thrombi of CE and LAA origin were identified. The results are useful for understanding the pathogenesis and for secondary stroke prevention.