Multinomial machine learning identifies independent biomarkers by integrated metabolic analysis of acute coronary syndrome

Abstract Background Small-molecule metabolite variations may reflect etiologies of acute coronary syndrome (ACS) and serve as biomarkers of ACS. Major confounders may exert spurious effects on the relationship between metabolism and ACS. It aims to identify independent biomarkers for different types of ACS by integrating of serum and urinary metabolomics. Methods We performed liquid chromatography-tandem mass spectrometry (LC-MS/MS)-based metabolomics study on serum and urine samples from 44 patients with unstable angina (UA), 77 with acute myocardial infarction (AMI), and 29 healthy controls (HC). Multinomial machine-learning-based integrated metabolite profiling and assessment of the confounders were used to integrate a biomarker panel for distinguishing the three groups. Results Different metabolic landscapes were portrayed for HC vs. UA, HC vs. AMI, and UA vs. AMI. Specifically, ACS risk was associated with metabolites increasing in alanine, aspartate and glutamate metabolism, D-glutamine and D-glutamate metabolism, and butanoate metabolism. An integrated model dependent on ACS, including 2-ketobutyric acid, SM (d18:1/20:0) of serum, and argininosuccinic acid, N6-Acetyl-L-lysine of urine, demarcated different ACS patients, providing a C-index of 0.993 (HC vs. UA), 0.941 (HC vs. AMI), and 0.930 (UA vs. AMI). Moreover, the four metabolites dynamically altered with ACS severity and positively or negatively correlated with ACS phenotypes. Conclusion The integration of serum and urinary metabolites provided an independent diagnostic biomarker panel for ACS.