Gene Correlation Network Analysis to Identify Biomarkers of Peri-Implantitis

Abstract

Background and Objectives: The histopathological and clinical conditions for transforming peri-implant mucositis into peri-implantitis (PI) are not fully clarified. We aim to uncover molecular mechanisms and new potential biomarkers of PI. Materials and Methods: Raw GSE33774 and GSE57631 datasets were obtained from the Gene Expression Omnibus (GEO) database. The linear models for microarray data (LIMMA) package in R software completes differentially expressed genes (DEGs). We conducted a weighted gene co-expression network analysis (WGCNA) on the top 25% of altered genes and identified the key modules associated with the clinical features of PI. Gene Ontology (GO) enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were performed using the R software. We constructed a protein-protein interaction (PPI) network through the STRING database. After that we used Cytohubba plug-ins of Cytoscape to screen out the potential hub genes, which were subsequently verified via receiver operating characteristic (ROC) curves in another dataset, GSE178351, and revalidation of genes through the DisGeNET database. Results: We discovered 632 DEGs (570 upregulated genes and 62 downregulated genes). A total of eight modules were screened by WGCNA, among which the turquoise module was most correlated with PI. The Cytohubba plug-ins were used for filtering hub genes, which are highly linked with PI development, from the candidate genes in the protein-protein interaction (PPI) network. Conclusions: We found five key genes from PI using WGCNA. Among them, ICAM1, CXCL1, and JUN are worthy of further study of new target genes, providing the theoretical basis for further exploration of the occurrence and development mechanism of PI.

Keywords: biomarkers; peri-implantitis; weighted gene co-expression network analysis (WGCNA).

Figure 1

DEGs based on the GSE57631. (A) The volcano diagram of up and down DEGs. Grey dots represent genes that are not differentially expressed or statistically significant. (B) The heatmap of the top 10 significantly up− or down-regulated genes with the largest differential multiples. (Abbreviations: DEGs, differentially expressed genes; GSE, GEO Series; GEO, Gene Expression Omnibus).

Figure 2

WGCNA diagram. (A) Sample clustering. (GSM835237 was removed from the queue) (B,C) Selection of soft-threshold β. (D) Clustering dendrogram. (E) Heatmap of the correlation between modules and clinical characteristics. (F) PI gene significance vs. membership in the turquoise module is shown in a scatter plot. (G) Venn diagram for screening CGs. (Abbreviations: CGs, common genes).

Figure 2

WGCNA diagram. (A) Sample clustering. (GSM835237 was removed from the queue) (B,C) Selection of soft-threshold β. (D) Clustering dendrogram. (E) Heatmap of the correlation between modules and clinical characteristics. (F) PI gene significance vs. membership in the turquoise module is shown in a scatter plot. (G) Venn diagram for screening CGs. (Abbreviations: CGs, common genes).

Figure 3

GO enrichment and KEGG analyses. The plot for the top 3 terms of GO enrichment and KEGG analyses. (Abbreviations: GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; BP, biological process; CC, cellular component; MF, molecular function).

Figure 4

PPI network. (A) 51 crucial common genes in PPI network. (B) Venn diagram displaying the intersection of five algorithms in Cytohubba with the top 10 genes (EPC, MNC, MCC, Closeness, and Radiality). (C) Top 5 hub genes were predicted by the STRING database. (Abbreviations: PPI, protein–protein interaction; MCC, Maximal Clique Centrality; EPC, Edge Percolated Component; MNC, Maximum Neighborhood Component).

Figure 5

Validation of hub genes. (A) Predicted ROC curves of hub genes based on GSE178351. (B) Venn diagram of intersecting genes of top 5 genes and PI-related genes from the DisGeNET database.