Screening of periodontitis-related diagnostic biomarkers based on weighted gene correlation network analysis and machine algorithms

BACKGROUND: Periodontitis is a common oral immune inflammatory disease and early detection plays an important role in its prevention and progression. However, there are no accurate biomarkers for early diagnosis. OBJECTIVE: This study screened periodontitis-related diagnostic biomarkers based on weighted gene correlation network analysis and machine algorithms. METHODS: Transcriptome data and sample information of periodontitis and normal samples were obtained from the Gene Expression Omnibus (GEO) database, and key genes of disease-related modules were obtained by bioinformatics. The key genes were subjected to Gene Ontology (GO) enrichment analysis, Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis and 5 machine algorithms: Logistic Regression (LR), Random Forest (RF), Gradient Boosting Decisio Tree (GBDT), Extreme Gradient Boosting (XGBoost), and Support Vector Machine (SVM). Expression and correlation analysis were performed after screening the optimal model and diagnostic biomarkers. RESULTS: A total of 47 candidate genes were obtained, and the LR model had the best diagnostic efficiency. The COL15A1, ICAM2, SLC15A2, and PIP5K1B were diagnostic biomarkers for periodontitis, and all of which were upregulated in periodontitis samples. In addition, the high expression of periodontitis biomarkers promotes positive function with immune cells. CONCLUSION: COL15A1, ICAM2, SLC15A2 and PIP5K1B are potential diagnostic biomarkers of periodontitis.