Machine learning identifies novel coagulation genes as diagnostic and immunological biomarkers in ischemic stroke

Background: Coagulation system is currently known associated with the development of ischemic stroke (IS). Thus, the current study is designed to identify diagnostic value of coagulation genes (CGs) in IS and to explore their role in the immune microenvironment of IS. Methods: Aberrant expressed CGs in IS were input into unsupervised consensus clustering to classify IS subtypes. Meanwhile, key CGs involved in IS were further selected by weighted gene co-expression network analysis (WGCNA) and machine learning methods, including random forest (RF), support vector machine (SVM), generalized linear model (GLM) and extreme-gradient boosting (XGB). The diagnostic performance of key CGs were evaluated by receiver operating characteristic (ROC) curves. At last, quantitative PCR (qPCR) was performed to validate the expressions of key CGs in IS. Results: IS patients were classified into two subtypes with different immune microenvironments by aberrant expressed CGs. Further WGCNA, machine learning methods and ROC curves identified ACTN1, F5, TLN1, JMJD1C and WAS as potential diagnostic biomarkers of IS. In addition, their expressions were significantly correlated with macrophages, neutrophils and/or T cells. GSEA also revealed that those biomarkers may regulate IS via immune and inflammation. Moreover, qPCR verified the expressions of ACTN1, F5 and JMJD1C in IS. Conclusions: The current study identified ACTN1, F5 and JMJD1C as novel coagulation-related biomarkers associated with IS immune microenvironment, which enriches our knowledge of coagulation-mediated pathogenesis of IS and sheds light on next-step in vivo and in vitro experiments to elucidate the relevant molecular mechanisms.