Integrating multi-omics and machine learning strategies to explore the "gene-protein-metabolite" network in ischemic heart failure with Qi deficiency and blood stasis syndrome

Abstract

Background: Ischemic heart failure (IHF) is a multifaceted syndrome associated with significant mortality and high hospitalization rates globally. According to traditional Chinese medicine (TCM) theory, Qi Deficiency and Blood Stasis (QXXY) Syndrome serves as the pathological basis of IHF. This study aims to investigate the biological basis of QXXY syndrome in IHF patients through an integrated multi-omics approach.

Methods: We enrolled 100 participants, comprising 40 IHF patients with QXXY syndrome (IHF-QXXY), 40 IHF patients without QXXY syndrome, and 20 healthy controls. Utilizing an integrated approach combining RNA sequencing (RNA-seq), data-independent acquisition (DIA) proteomics, and targeted metabolomics, we established a comprehensive "gene-protein-metabolite" network for IHF-QXXY syndrome. Candidate biomarkers were identified through machine learning algorithms and further validated using RT-qPCR and targeted proteomics via intelligent parallel reaction monitoring (iPRM).

Results: Patients with IHF-QXXY syndrome present with pronounced disruptions in energy metabolism, chronic inflammation, and coagulation abnormalities. The "gene-protein-metabolite" network of IHF-QXXY syndrome comprises six mRNAs, four proteins, and five metabolites. Key pathways involve the activation of neutrophil extracellular traps formation, platelet activation, the HIF-1 signaling pathway, and glycolysis/gluconeogenesis, alongside the suppression of the citrate cycle and oxidative phosphorylation. The key metabolites potentially associated with QXXY syndrome include 3-methylpentanoic acid, arachidonic acid, N-acetylaspartylglutamic acid, L-acetylcarnitine, and 12-hydroxystearic acid. We identified a panel of candidate biomarkers, including HIF-1α, IL10, PAD4, ACTG1, SOD2, GAPDH, FGA, FN1, F13A1, and ATP5PF. This biomarker combination significantly enhanced the diagnostic performance of IHF-QXXY syndrome (AUC > 0.863) and retained high diagnostic accuracy during validation (AUC > 0.75).

Conclusion: This study provides a comprehensive characterization of the molecular features of QXXY syndrome in IHF patients, highlighting key pathways and biomarkers linked to energy metabolism dysregulation, chronic inflammation, and coagulation abnormalities. These findings may provide novel insights and methods for further advancing this research field.

Keywords: Ischemic heart failure; Machine learning; Molecular docking; Multi-omics; Traditional Chinese medicine; Yiqi Huoxue.

Fig. 1

Biological characteristics of IHF-QXXY syndrome. A The level of ATP in each group was detected by ELISA kit. B The level of Acetyl-CoA in each group was detected by ELISA kit. C The level of ET-1 in each group was detected by ELISA kit. D The level of NO in each group was detected by ELISA kit. E The level of ICAM-1 in each group was detected by ELISA kit. F The level of VCAM-1 in each group was detected by ELISA kit. G The level of TNF-α in each group was detected by ELISA kit. H The level of IL-1β in each group was detected by ELISA kit. I The level of IL-6 in each group was detected by ELISA kit. J The level of PGI2 in each group was detected by ELISA kit. K The level of TAT in each group was detected by ELISA kit. L The level of TXA2 in each group was detected by ELISA kit. M The level of MPO in each group was detected by ELISA kit. N The level of cfDNA in each group was detected by ELISA kit. O The level of Cit-H3 in each group was detected by ELISA kit. P The level of NE in each group was detected by ELISA kit. *P < 0.05; ** P < 0.01; *** P < 0.001; **** P < 0.0001

Fig. 2

Transcriptomic characteristics of IHF-QXXY syndrome. A Classification results of HC, IHF-QXXY, and IHF-NQXXY groups based on PCA score plot. B The volcano plots of DEGs between IHF-QXXY and HC group. C The volcano plots of DEGs between IHF-NQXXY and HC group. D Venn diagram of specific DEGs to IHF-QXXY syndrome. E Pathway enrichment of specific DEGs to IHF-QXXY syndrome. F PPI network analysis of specific DEGs to IHF-QXXY syndrome. G The top 17 nodes of the PPI network identified by the MCC, MNC, and Degree algorithms. H LASSO analysis screening of feature genes. I SVM-RFE algorithm screening of feature genes. J RF algorithm screening of feature genes. K ROC curve of HIF-1α, IL10, PAD4, ACTG1, SOD2, and GAPDH. L Combined diagnostic ROC curves of feature genes

Fig. 3

Proteomic characteristics of IHF-QXXY syndrome. A Classification results of HC, IHF-QXXY, and IHF-NQXXY groups based on PLS-DA score plot. B The volcano plots of DEPs between IHF-QXXY and HC group. C The volcano plots of DEPs between IHF-NQXXY and HC group. D Venn diagram of specific DPGs to IHF-QXXY syndrome. E Pathway enrichment of specific DEPs to IHF-QXXY syndrome. F PPI network analysis of specific DEPs to IHF-QXXY syndrome. G The top 19 nodes of the PPI network identified by the MCC, MNC, and Degree algorithms. H LASSO analysis screening of feature proteins. I SVM-RFE algorithm screening of feature proteins. J RF algorithm screening of feature proteins. K ROC curve of FGA, FN1, F13A1, and ATP5PF. L Combined diagnostic ROC curves of feature proteins

Fig. 4

Metabolomic characteristics of IHF-QXXY syndrome. A Pie chart of identified metabolites. B PLS-DA score plot of the HC, IHF-QXXY, and IHF-NQXXY groups. C The volcano plots of DMs between IHF-QXXY and HC group. D The volcano plots of DMs between IHF-NQXXY and HC group. E Venn diagram of specific DMs to IHF-QXXY syndrome. F Pathway enrichment of specific DMs to IHF-QXXY syndrome. G ROC curve of 3-methylpentanoic acid, arachidonic acid, N-acetylaspartylglutamic acid, L-acetylcarnitine, and 12-hydroxystearic acid. H Combined diagnostic ROC curves of Key metabolites

Fig. 5

Integrated analysis of multi-omics experiments. A ClueGO analysis of KEGG pathway enrichment. B Enrichment analysis of KEGG pathway shared by DEGs, DEPs, and DMs. C IHF-QXXY syndrome “gene-protein-metabolite” network

Fig. 6

The expression of candidate markers was verified by RT-qPCR and iPRM. A The mRNA expression of ACTG1. B The mRNA expression of HIF-1α. C The mRNA expression of PAD4. D The mRNA expression of SOD2. E The mRNA expression of GAPDH. F The mRNA expression of IL-10. G iPRM validation of key DEPs. H Validated ROC curves of HIF-1α, IL10, PAD4, ACTG1, SOD2, and GAPDH. I Validated ROC curve of FGA, FN1, F13A1, and ATP5PF. ^*P < 0.05; ^** P < 0.01