Discovery and Validation of Potential Serum Biomarkers for Heart Failure by Untargeted Metabolomics

Abstract

Detection of biomarkers was extremely important for the early diagnosis, prognosis, and therapy optimization of diseases. The purpose of this study was to investigate the differences in serum metabolites between patients with heart failure (HF) and healthy control (HC) and to diagnose HF qualitatively. In this study, serum samples from 83 patients with HF and 35 HCs were used as the research subjects for untargeted metabolomic analysis using ultraperformance liquid chromatography combined with quadrupole-time of flight mass spectrometry (UPLC-QTOF/MS) technology. Potential biomarkers were screened and validated using the orthogonal partial least squares discriminant analysis (OPLS-DA), random forest (RF), binary logistic regression (BLR), and receiver operating characteristic (ROC) analysis. The results indicated that a total of 43 metabolites were considered as differentially expressed metabolites (DEMs). Among these DEMs, glycodeoxycholate was identified as a specific biomarker of HF. A ROC curve analysis for HC versus HF discrimination showed an area under the ROC curve (AUC) of 0.9853 (95% CI: 0.9859-1.0000), a sensitivity of 95%, and a specificity of 100%. Hence, glycodeoxycholate might serve as a potential biomarker for HF. Furthermore, the amino acid metabolism was screened as the most significantly altered pathway in patients with HF. By identifying serum biomarkers and analyzing metabolic pathways, our study provided opportunities to enhance the understanding of the pathogenesis and early diagnosis of HF.

Keywords: UPLC/Q-TOF-MS; biomarker; heart failure; metabolite profile; serum metabolomics.

Figure 1

Workflow of this study.

Figure 2

Identification of potential metabolic biomarkers for the diagnosis of HF in the discovery set and validation set. (a) OPLS-DA score plot based on HC and HF groups in the discovery set. (b) Two hundred permutation tests of OPLS-DA model in the discovery set. (c) OPLS-DA score plot based on HC and HF groups in the validation set. (d) Two hundred permutation tests of OPLS-DA model in the discovery set.

Figure 3

(a) Mean decrease accuracy values of the metabolic biomarkers used for random forest classification. (b) Heat map of 24 differential metabolites. (c) Pathway analysis of the differentially altered metabolites identified in patients with HF.

Figure 4

(a) ROC curve analysis of glycodeoxycholate in the discovery set. (b) ROC curve analysis of glycodeoxycholate in the validation set. (c) Correlation analysis between the glycodeoxycholate and LVEF. (d) Scatter plot of the relative intensity of glycodeoxycholate.