Identification of risk genes related to myocardial infarction and the construction of early SVM diagnostic model

Background: Myocardial Infarction (MI) is a fatal cardiovascular system disease. At present, the diagnosis of MI patients is mainly based on the patient's clinical manifestations, dynamic changes in electrocardiogram (ECG), and changes in myocardial enzymes. ECG is insufficient to diagnose an acute coronary syndrome or acute myocardial infarction, since ST-segment deviation might be also present in other conditions, such as acute pericarditis and early repolarization patterns. Given the low specificity and effectiveness of the current diagnostic strategies, an accurate diagnostic approach based on the level of gene expression is urgently needed in the clinic.

Methods and results: We compared the gene's expression between MI patients and normal samples. The RNAseq data were downloaded from the GEO database. Differentially expressed genes underwent a feature selection process, and the signatures were selected to train a machine-learning model. In this study, we identified the risk genes associated with MI as signatures and uses the SVM to establish a diagnostic model. The accuracy of the model on discovery data is 0.87, which significantly improves the diagnostic efficiency of early detection of MI patients (MIPs). Two independent datasets were applied to verify the diagnostic model. Our model can effectively distinguish the control group from the disease group.

Conclusions: We used risk genes to construct a diagnostic model for MI diagnosis, which can effectively distinguish MIPs from normal samples in the both of the discovery data and validation data. In the validation data, we found that percutaneous coronary intervention could indeed reverse MI to a certain extent, and the gene expression level of patients treated with percutaneous coronary intervention (PCI) was closer to the normal state.