Comprehensive bioinformatic methods and machine learning algorithms to identify diagnostic biomarkers of coronary artery disease progression in pulmonary arterial hypertension

Abstract

Abstract Background Numerous fundamental and clinical investigations have showcased the correlation and interplay mechanism between coronary artery disease (CAD) and pulmonary arterial hypertension (PAH). We aimed to investigate diagnostic indicators and the correlation between immune response and diagnostic indicators.Methods To conduct additional differential expression analysis and weighted gene coexpression network analysis (WGCNA), we obtained CAD and PAH data from the Gene Expression Omnibus (GEO) database. Next, we employed shared genes to conduct enrichment analysis, construct a protein-protein interaction (PPI) network, and subsequently identify diagnostic biomarkers through the utilization of three machine learning algorithms. The diagnostic biomarkers were utilized for conducting logistic regression analysis and constructing a nomogram. Then, we compared the expression differences and their respective diagnostic effects. The evaluation of immune infiltration was conducted finally.Results By intersecting 671 genes that were differentially expressed in CAD and 2052 genes that were key module genes in PAH, we identified 67 genes that were common to both conditions. These shared genes were mainly enriched in signaling pathways associated with the activation of leukocytes and the regulation of inflammation. We further identified 26 genes through PPI network construction. Afterwards, three machine learning algorithms were utilized to choose two candidate biomarkers, namely DPYD and CPT1A. The two possible indicators showed improved diagnostic effectiveness and suggested a statistically significant positive correlation with macrophages in individuals with progressive CAD.Conclusion In this study, we have conducted the first research to identify early diagnostic biomarkers for the advancement of CAD in association with PAH. This was achieved by employing a variety of bioinformatic methods and machine learning algorithms. Possible therapeutic targets may exist due to the presence of a favorable correlation between diagnostic biomarkers and immune cells.