Identification and validation of novel marker genes to predict potential gestational diabetes mellitus patients by WGCNA and machine learning.

Abstract

Objectives: To identify novel marker genes to predict potential gestational diabetes mellitus (GDM) patients.

Material and: METHODS: Based on Gene Expression Omnibus (GEO) datasets, the differentially expressed genes (DEGs) between control and GDM were identified, followed by enrichment analysis and protein-protein interaction (PPI) network construction. Then, Weighted gene co-expression network analysis (WGCNA) was conducted to screen the key module genes, then the important genes were obtained. In addition, Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression, Support Vector Machine - Recursive Feature Elimination (SVM-RFE), and random forest (RF) were employed to identify the key genes. Receiver operating characteristic (ROC) analysis was performed to assess the diagnostic efficacy of key genes, and a nomogram was developed. The correlation between key genes and immune cells was analyzed, and miRNA-mRNA-TF network was constructed.

Results: A total of 257 DEGs were screened between control and GDM groups, and these DEGs were involved in p53 signaling pathway, cell cycle and oocyte meiosis pathways. Then PPI network was constructed, including 163 nodes and 5502 interaction relationships. After WGCNA and machine learning, a total of 4 key genes were obtained, including SNRPD3, NGDN, ANKRD36 and TAS2R20, followed by a nomogram was constructed. SNRPD3 was positively correlated with CD8 T cells. miRNA-mRNA-TF network was conducted, including 56 miRNAs, 4 mRNAs, and 32 TFs. Besides, luteolin PC3 UP, alsterpaullone PC3 UP, and solanine HL60 UP were associated with NGDN, and MeIQx CTD 00001739 was related to TAS2R20.

Conclusions: Four key marker genes for predicting potential GDM were identified, including SNRPD3, NGDN, ANKRD36 and TAS2R20, and a nomogram was established for predicting potential GDM patients.