Targeted Quantitative Metabolomics and Lipidomics Reveal Dysregulated Metabolic Networks and a Serum Candidate Biomarker for Atrial Fibrillation

Figures

• 

  Figure 1. Schematic workflow of targeted serum metabolomic and lipidomic profiling for atrial fibrillation (AF) research. Serum samples were collected from AF patients and healthy controls and analyzed via alternating reversed-phase liquid chromatography–mass spectrometry (RPLC-MS/MS). Multivariate statistics, univariate analysis, and multiscale embedded correlation network analysis (MECNA) were performed to screen differential metabolites and core hub molecules. Binary logistic regression was applied to construct and validate a two-molecule diagnostic biomarker panel with receiver operating characteristic (ROC) curve evaluation.

• 

  Figure 2. Data quality control and classification distribution of serum metabolites and lipids. (A,B) Relative standard deviation (RSD) distribution of 218 metabolites and 528 lipids from quality control (QC) samples to assess analytical stability. (C,D) Pie charts illustrating the subclasses and proportional distribution of qualified metabolites and lipids after filtering. Abbreviations: LPCs, lysophosphatidylcholines; LPEs, lysophosphatidylethanolamines; CEs, cholesterol esters; Cers, ceramides; HexCers, exosylceramides; Hex2Cers, dihexaglycosylceramides; DGs, diacylglycerols; SMs, sphingomyelins; TGs, triacylglycerols; FAs, fatty acids; PIs, phosphatidylinositols; PCs, phosphatidylcholines; PEs, phosphatidylethanolamines.

• 

  Figure 3. Global metabolic alterations in serum from AF patients versus healthy controls. (A,B) PCA score plots of metabolome and lipidome datasets showing distinct separation between AF and control cohorts. (C,D) Volcano plots displaying differential metabolites and lipids. Vertical dashed lines mark log2(FC) corresponding to fold-change thresholds of 0.67 (downregulated) and 1.5 (upregulated), and the horizontal dashed line denotes −log10(FDR) = 2 (FDR = 0.01). Molecules with false discovery rate (FDR) < 0.05 and fold change (FC) > 1.5 (upregulated) or FC < 0.67 (downregulated) are highlighted in colored dots.

• 

  Figure 4. Multiscale embedded correlation network analysis (MECNA) of AF-associated differential metabolites and lipids. (A) Global correlation network constructed from 312 significantly AF-related molecules, with nodes colored according to metabolite/lipid subclasses and key hub metabolites labeled. (B) Subnetwork centered on core hub fatty acids FA 20:3 and FA 20:2. (C) Heatmap showing relative serum abundance of molecules in the FA-centered subnetwork between AF patients and healthy controls.

• 

  Figure 5. Diagnostic performance of the dual-molecule biomarker panel (Phe-Trp + FA 22:5) constructed via binary logistic regression. (A,B) ROC curves and corresponding area under the curve (AUC) values in the discovery cohort and independent validation cohort. (C,D) Scatter plots displaying predicted AF probabilities of all participants in discovery and validation cohorts based on the logistic regression model. Vertical red dashed lines indicate the cutoff obtained for group classification.