Abstract
Cardiovascular disease (CVD) represents one of the main causes of mortality worldwide and nearly a half of it is related to ischemic heart disease (IHD). The article represents a comprehensive study on the diagnostics of IHD through the targeted metabolomic profiling and machine learning techniques. A total of 112 subjects were enrolled in the study, consisting of 76 IHD patients and 36 non-CVD subjects. Metabolomic profiling was conducted, involving the quantitative analysis of 87 endogenous metabolites in plasma. A novel regression method of age-adjustment correction of metabolomics data was developed. We identified 36 significantly changed metabolites which included increased cystathionine and dimethylglycine and the decreased ADMA and arginine. Tryptophan catabolism pathways showed significant alterations with increased levels of serotonin, intermediates of the kynurenine pathway and decreased intermediates of indole pathway. Amino acid profiles indicated elevated branched-chain amino acids and increased amino acid ratios. Short-chain acylcarnitines were reduced, while long-chain acylcarnitines were elevated. Based on these metabolites data, machine learning algorithms: logistic regression, support vector machine, decision trees, random forest, and gradient boosting, were used for IHD diagnostic models. Random forest demonstrated the highest accuracy with an AUC of 0.98. The metabolites Norepinephrine; Xanthurenic acid; Anthranilic acid; Serotonin; C6-DC; C14-OH; C16; C16-OH; GSG; Phenylalanine and Methionine were found to be significant and may serve as a novel preliminary panel for IHD diagnostics. Further studies are needed to confirm these findings.