ObjectiveTo apply different machine learning menthods in screening of core autophagy genes in vulnerable plaques and investigate their association with immune infiltration.MethodsStable and unstable plaque samples from gene expression omnibus (GEO) datasets (GSE163154, GSE41571, GSE111782) were integrated, and through intersection of differentially expressed genes with the Human Autophagy Database, 14 candidate genes were obtained. Core autophagy genes were screened using least absolute shrinkage and selection operator (LASSO) regression and support vector machine recursive feature elimination (SVM-RFE). Diagnostic value of core autophagy genes for unstable plaques were evaluated by receiver operating characteristic (ROC) curve analysis. A nomogram model was constructed to analyze the predictive value of core autophagy genes for unstable plaques, the predictive efficacy was verified using calibration curves and decision analysis curves. The CIBESORT algorithm was used to analyze the immune infiltration characteristics in stable plaque samples and unstable plaque samples. Correlations between core autophagy genes and different immune cells was analyzed by Pearson's method.ResultsFour core autophagy genes (CX3CL1, CTSD, MTMR14, and NRG1) were identified using two different machine learning methods. ROC curve analysis demonstrated their diagnostic value for unstable plaques, with area under the curve (AUC) values of 0.874, 0.876, 0.866, and 0.733, respectively. Calibration curve analysis indicated strong predictive performance of the nomogram model based on these core autophagy genes, showing high concordance with the ideal curve. Decision analysis curve confirmed favorable net benefits for predicting unstable plaques using the core autophagy gene nomogram model. Immune infiltration analysis revealed significant differences in immune cell composition between stable and unstable plaques and unstable plaques exhibited increased M0 macrophage infiltration (P < 0.05), but with a non-significant decreasing trend in M2 macrophages. The expression level of core autophagy genes had a strong correlation with the infiltration degree of various immune cells. The expression of some core autophagy genes showed significant correlations with M1 macrophage (P < 0.05).ConclusionMachine learning identified four core autophagy genes (CX3CL1, CTSD, MTMR14, and NRG1) that demonstrate significant diagnostic and predictive value for unstable plaques. The expression levels of these core autophagy genes are correlated with infiltration of macrophage.