Identification of potential diagnostic biomarkers for myocardial infarction using bioinformatics and machine learning

Abstract

Acute myocardial infarction(AMI)is an important type of cardiovascular disease, which seriously threatens the lives of humans. In order to help patients receive timely clinical treatment and improve their survival rates, it is necessary to screen out related molecules in advance. To this end, we applied bioinformatics methods and machine learning algorithms to find possible biomarkers associated with AMI. When considering possible connections with dynamics of immune cells activation. Based on the RNA-seq results, differential expression analysis was performed followed by WGCNA to identify stably changed genes. The KEGG and GO enrichment analyses revealed that the identified genes are mainly related with inflammatory response, immune processes, apoptosis and immunomodulation. In our PPI network, some hub genes (e.g., FOS, JUN, TNF, IL1B, TLR2) were located in the center of the whole network. We found strong associations between the major genes and their interaction effects, we developed a predictive model for diagnosis using an artificial intelligence algorithm as follows: suggesting a substantially increased discriminative performance for AMI with the 4 hub genes NFKBIA, FCER1G, CD36 and ICAM1. The model achieved a high AUC score of 0.924.

External validation based on an independent single-cell sequencing dataset demonstrated that NFKBIA, FCER1G, and ICAM1 displayed consistent expression patterns in both the training and test cohorts, with the highest expression levels observed in cardiomyocytes. Accordingly, these three genes may serve as reliable biomarkers for AMI and are closely associated with the distribution of immune cells.