Identification of biomarkers associated with programmed cell death in liver ischemia-reperfusion injury: insights from machine learning frameworks and molecular docking in multiple cohorts

Abstract

Introduction: Liver ischemia-reperfusion injury (LIRI) is a major reason for liver injury that occurs during surgical procedures such as hepatectomy and liver transplantation and is a major cause of graft dysfunction after transplantation. Programmed cell death (PCD) has been found to correlate with the degree of LIRI injury and plays an important role in the treatment of LIRI. We aim to comprehensively explore the expression patterns and mechanism of action of PCD-related genes in LIRI and to find novel molecular targets for early prevention and treatment of LIRI.

Methods: We first compared the expression profiles, immune profiles, and biological function profiles of LIRI and control samples. Then, the potential mechanisms of PCD-related differentially expressed genes in LIRI were explored by functional enrichment analysis. The hub genes for LIRI were further screened by applying multiple machine learning methods and Cytoscape. GSEA, GSVA, immune correlation analysis, transcription factor prediction, ceRNA network analysis, and single-cell analysis further revealed the mechanisms and regulatory network of the hub gene in LIRI. Finally, potential therapeutic agents for LIRI were explored based on the CMap database and molecular docking technology.

Results: Forty-seven differentially expressed genes associated with PCD were identified in LIRI, and functional enrichment analysis showed that they were involved in the regulation of the TNF signaling pathway as well as the regulation of hydrolase activity. By utilizing machine learning methods, 11 model genes were identified. ROC curves and confusion matrix from the six cohorts illustrate the superior diagnostic value of our model. MYC was identified as a hub PCD-related target in LIRI by Cytoscape. Finally, BMS-536924 and PF-431396 were identified as potential therapeutic agents for LIRI.

Conclusion: This study comprehensively characterizes PCD in LIRI and identifies one core molecule, providing a new strategy for early prevention and treatment of LIRI.

Keywords: biomarkers; liver ischemia–reperfusion injury; machine learning; molecular docking; programmed cell death.

Figure 1

The process of data analyzing in this study.

Figure 2

Analysis of differences between LIRI and control samples. (A) The volcano plot for DEGs in LIRI; (B) Differences in immune cells between LIRI and controls; (C–F) GSEA analysis of KEGG and GO between LIRI and controls.

Figure 3

Functional enrichment analysis of PCD-related DEGs. (A) The intersection of DEGs and PCD-related gene; (B) PPI of the PCD-related DEGs; (C) GeneMANIA analysis of PCD-related DEGs; (D) Functional enrichment analyses by the Metascape database.

Figure 4

The model genes of LIRI were screened by machine learning method. (A) AUC values of 113 machine learning models; (B) ROC curves of diagnostic performance of 11 model genes for LIRI; (C–F) Confusion matrix of the model in training and three test sets.

Figure 5

Identification of the hub gene. (A) Box plot showed the expression difference of model genes between LIRI and control samples in training set. (B) Correlation analysis of 11 model genes; (C) PPI of the model genes; (D) The hub gene MYC was identified from 11 genes using Cytoscape; (E) Immune cell correlation analysis of MYC.

Figure 6

Functional enrichment analysis of MYC in LIRI. (A,B) Enrichment biological functions and pathways of MYC identified by GSVA; (C–F) Enrichment biological functions and pathways of MYC identified by GSEA.

Figure 7

Regulatory network, single-cell mapping and immunofluorescence analysis of MYC. (A) Prediction of TFs based on different databases; (B) The ceRNA network of MYC; (C) The immunofluorescence of MYC based on HPA database; (D) The single-cell mapping of MYC based on HPA database; (E) Molecular docking of BMS-536924 and MYC; (F) Molecular docking of PF-431396 and MYC.