Construction of the coexpression network involved in the pathogenesis of thyroid eye disease via bioinformatics analysis

Abstract

Background: Thyroid eye disease (TED) is the most common orbital pathology that occurs in up to 50% of patients with Graves' disease. Herein, we aimed at discovering the possible hub genes and pathways involved in TED based on bioinformatical approaches.

Results: The GSE105149 and GSE58331 datasets were downloaded from the Gene Expression Omnibus (GEO) database and merged for identifying TED-associated modules by weighted gene coexpression network analysis (WGCNA) and local maximal quasi-clique merger (lmQCM) analysis. EdgeR was run to screen differentially expressed genes (DEGs). Transcription factor (TF), microRNA (miR) and drug prediction analyses were performed using ToppGene suite. Function enrichment analysis was used to investigate the biological function of genes. Protein-protein interaction (PPI) analysis was performed based on the intersection between the list of genes obtained by WGCNA, lmQCM and DEGs, and hub genes were identified using the MCODE plugin. Based on the overlap of 497 genes retrieved from the different approaches, a robust TED coexpression network was constructed and 11 genes (ATP6V1A, PTGES3, PSMD12, PSMA4, METAP2, DNAJA1, PSMA1, UBQLN1, CCT2, VBP1 and NAA50) were identified as hub genes. Key TFs regulating genes in the TED-associated coexpression network, including NFRKB, ZNF711, ZNF407 and MORC2, and miRs including hsa-miR-144, hsa-miR-3662, hsa-miR-12136 and hsa-miR-3646, were identified. Genes in the coexpression network were enriched in the biological processes including proteasomal protein catabolic process and proteasome-mediated ubiquitin-dependent protein catabolic process and the pathways of endocytosis and ubiquitin-mediated proteolysis. Drugs perturbing genes in the coexpression network were also predicted and included enzyme inhibitors, chlorodiphenyl and finasteride.

Conclusions: For the first time, TED-associated coexpression network was constructed and key genes and their functions, as well as TFs, miRs and drugs, were predicted. The results of the present work may be relevant in the treatment and diagnosis of TED and may boost molecular studies regarding TED.

Keywords: Autoimmune inflammatory disease; Differentially expressed genes; Gene coexpression networks; Protein–protein interaction; Thyroid eye disease.

Fig. 1

Data preprocessing and differential expression gene analysis. Boxplot of the merged matrix of transcriptome data A before and B after batch effect removal and normalization. C Volcano plot of DEGs. Top 10 DEGs (sorted by adj. p value) in the up-regulated and down-regulated groups. D Heatmap of DEGs. Top 10 DEGs (sorted by |log2FC|) in the up-regulated and down-regulated groups

Fig. 2

Construction of WGCNA networks. A Sample dendrogram and trait heatmap. The two traits are TED and normal. B Scale independence and mean connectivity of various soft-thresholding values (β). C Gene dendrogram and modules color. D Eigengene adjacency heatmap. E The heatmap of the 400 genes in the coexpression network

Fig. 3

WGCNA and lmQCM modules and functional analysis of key modules. A Module–trait relationships in WGCNA modules. B Function enrichment analysis of the blue module most significantly correlated with TED as identified by WGCNA. C Module–trait relationships in lmQCM modules. D Function enrichment analysis of the red module most significantly correlated with TED as identified by lmQCM

Fig. 4

The overlap of DEGs and genes in key TED-associated modules identified by WGCNA and lmQCM. A The Venn plot of the DEGs and genes in key TED-associated modules identified by WGCNA and lmQCM. B The function enrichment analysis of the overlap of DEGs and genes in key TED-associated modules identified by WGCNA and lmQCM

Fig. 5

Construction of a robust TED coexpression network based on the intersection genes. The 497 coexpression genes obtained from the intersection of the overlap of DEGs and genes in key TED-associated modules identified by WGCNA and lmQCM were imported in stringdb for protein–protein interaction (PPI) network, and the PPI network was downloaded

Fig. 6

Visualization of the coexpression network in Cytoscape and identification of clusters and hub genes. The PPI network of the TED coexpression genes was visualized in Cytoscape. Clusters were identified by MCODE, and genes with node degree higher than ten were visualized. Node labels in red indicated hub genes with node degree higher than ten

Fig. 7

Workflow to identify TED-specific coexpression modules and TED-associated pathway and hub genes