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

Jingjing Wei, Aolong Wang, Peng Yu, Yang Sun, Wenjun Wu, Yilin Zhang, Rui Yu, Bin Li, Mingjun ZhuChinese 2025 · DOI

TL;DR

This study integrates multi-omics data (genomics, proteomics, metabolomics) with machine learning to build a comprehensive molecular network for Qi Deficiency and Blood Stasis (QXXY) syndrome in Ischemic Heart Failure (IHF) patients. The findings reveal key pathways and biomarkers associated with energy metabolism, inflammation, and coagulation, offering potential for improved diagnostics and targeted therapies.

What Problem Does This Paper Address?

Ischemic Heart Failure (IHF) is a major global health burden with high mortality and hospitalization rates. Traditional Chinese Medicine (TCM) identifies "Qi Deficiency and Blood Stasis" (QXXY) syndrome as a key pathological basis for IHF. However, the precise molecular underpinnings of QXXY syndrome in IHF remain poorly understood, hindering its integration into modern medical practice and the development of targeted therapies. Existing research often relies on single-omics approaches, lacking depth and cross-validation. This paper aims to bridge this gap by comprehensively characterizing the molecular landscape of QXXY syndrome in IHF patients using an integrated multi-omics and machine learning strategy.

Methods

The researchers employed a multi-pronged approach:

  • Multi-omics Data Generation:
    • RNA-sequencing (RNA-seq): To analyze gene expression profiles from blood samples.
  • Data-Independent Acquisition (DIA) Proteomics: To quantify protein abundance in blood.
  • Targeted Metabolomics: To measure the levels of specific metabolites in blood.
  • Machine Learning for Biomarker Discovery:
    • LASSO Regression: Used for feature selection and regularization, helping to identify the most relevant genes/proteins.
  • Support Vector Machine Recursive Feature Elimination (SVM-RFE): Another feature selection method to identify a robust subset of biomarkers.
  • Random Forest (RF): Utilized for variable importance ranking to pinpoint key predictive features.
  • The intersection of features identified by these three methods was used to pinpoint candidate biomarkers.
  • Network Construction and Analysis:
    • Protein-Protein Interaction (PPI) Network: Built using the STRING database and visualized with Cytoscape to understand how identified genes and proteins interact.
  • KEGG Pathway Enrichment Analysis: Performed using ClueGO and CluePedia to identify the biological pathways significantly affected by the identified molecular changes.
  • Validation:
    • Real-time quantitative polymerase chain reaction (RT-qPCR): To validate the mRNA expression levels of selected candidate genes.
  • Intelligent Parallel Reaction Monitoring (iPRM): An advanced targeted proteomics technique to validate the abundance of selected candidate proteins.
  • Receiver Operating Characteristic (ROC) Curve Analysis: Used to assess the diagnostic performance of individual and combined biomarkers.

Datasets

The study utilized blood samples from 100 participants:

  • 40 IHF patients with QXXY syndrome (IHF-QXXY)
  • 40 IHF patients without QXXY syndrome (IHF-NQXXY)
  • 20 healthy controls (HC)

These samples were used for RNA-seq, DIA proteomics, and targeted metabolomics. A subset of 10 patients from each group was used for RT-qPCR validation.

Key Insights and Ideas

  • QXXY Syndrome as a Distinct Molecular Entity: The study successfully differentiated IHF-QXXY patients from IHF-NQXXY and healthy controls at the molecular level, suggesting QXXY syndrome represents a unique subtype of IHF with specific molecular alterations.
  • Energy Metabolism Dysregulation is Central: A core finding is the significant disruption in energy metabolism pathways, including downregulation of the citric acid cycle and oxidative phosphorylation, and upregulation of glycolysis. This aligns with TCM's "Qi deficiency" concept.
  • Inflammation and Coagulation as Key Players: The research highlights the critical roles of chronic inflammation (e.g., NETs formation, NF-kappa B signaling) and coagulation dysfunction (e.g., platelet activation) in the pathogenesis of QXXY syndrome.
  • Multi-omics Integration Unveils Synergies: Combining data from genomics, proteomics, and metabolomics provided a more holistic view, revealing how these different molecular layers interact to drive the disease phenotype.
  • Machine Learning Identifies Robust Biomarker Panels: The application of multiple machine learning algorithms effectively filtered down a large set of potential candidates to a concise panel of biomarkers with strong diagnostic potential.

Key Results

  • Biomarker Panels:
    • Transcriptomic Biomarkers: Six candidate biomarkers (HIF-1α, IL10, PAD4, ACTG1, SOD2, and GAPDH) were identified. A combined ROC analysis yielded an AUC of 0.863.
  • Proteomic Biomarkers: Four candidate biomarkers (FGA, FN1, F13A1, and ATP5PF) were identified. Their combined diagnostic performance achieved an AUC of 0.956.
  • Metabolomic Biomarkers: A panel of five metabolites (3-methylpentanoic acid, arachidonic acid, N-acetylaspartylglutamic acid, L-acetylcarnitine, and 12-hydroxystearic acid) showed enhanced diagnostic performance with an AUC of 0.866 when combined.
  • Pathway Involvement: Key pathways implicated include Neutrophil extracellular trap formation, Platelet activation, Complement and coagulation cascades, HIF-1 signaling pathway, Glycolysis/Gluconeogenesis, and Oxidative phosphorylation.
  • Clinical Differentiation: The IHF-QXXY group showed a higher LVEF and longer 6MWD compared to the IHF-NQXXY group, suggesting they might be in a relatively stable phase of IHF.

Limitations and Future Work

  • External Validation Needed: The authors acknowledge that the findings, while promising, require validation in independent, geographically and clinically distinct cohorts, ideally through multicenter collaborations.
  • Temporal Dynamics and Causality: The study identifies associations but cannot fully resolve temporal dynamics or establish causality. Future longitudinal studies and interventional trials are suggested to address this.
  • Functional Role Elucidation: The precise functional roles of the identified biomarkers and pathways in IHF pathogenesis need further investigation through cell and animal model studies.
  • TCM Integration: While the study bridges TCM theory with modern science, further work is needed to fully integrate these findings into clinical practice and develop TCM-based therapies for IHF.

Summary updated

Abstract

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.
BibTeX
@inproceedings{wei2025integrating,
  author = {Jingjing Wei and Aolong Wang and Peng Yu and Yang Sun and Wenjun Wu and Yilin Zhang and Rui Yu and Bin Li and Mingjun Zhu},
  title = {{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}},
  booktitle = {Chinese},
  year = {2025},
  doi = {10.1186/s13020-025-01151-9},
}
Figure