Identification of O-glycosylation related genes and subtypes in ulcerative colitis based on machine learning

Abstract

Ulcerative colitis (UC) is an immune-related inflammatory bowel disease, with its underlying mechanisms being a central area of clinical research. O-GlcNAcylation plays a critical role in regulating immunity progression and the occurrence of inflammatory diseases and tumors. Yet, the mechanism of O-GlcNAc-associated colitis remains to be elucidated. To this end, the transcriptional and clinical data of GSE75214 and GSE92415 from the GEO database was hereby examined, and genes MUC1, ADAMTS1, GXYLT2, and SEMA5A were found to be significantly related to O-GlcNAcylation using machine learning methods. Based on the four hub genes, two UC subtypes were built. Notably, subtype B might be prone to developing colitis-associated colorectal cancer (CAC). This study delved into the role of intestinal glycosylation changes, especially the O-GlcNAcylation, and forged a foundation for further research on the occurrence and development of UC. Overall, understanding the role of O-GlcNAcylation in UC could have significant implications for diagnosis and treatment, offering valuable insights into the disease's progression.

Fig 1. Identification and functional enrichment analysis of DEGs.

(A) The volcano map showed DEGs from two GEO datasets, UC and health control. (B) The heatmap showed the different genes between UC and healthy controls. The screening criteria were set to |LogFC| > 1 and adj.P.Val < 0.05. (C-E) The enrichment analysis results of GO, including BP, CC, and MF, revealed the underlying functions of DEGs. (F) KEGG revealed the first twenty pathways of differential gene enrichment.

Fig 2. Screening and functional enrichment of O-GlcNAcylation-associated differential genes.

(A) Venn diagram of upregulated differential genes and O- GlcNAcylation gene sets. (B) Venn diagram of downregulating differential gene and O-GlcNAcylation gene set. (C) The enrichment analysis results of GO, including BP, CC, and MF. (D) The KEGG enrichment of DEGs. (E) Mapping between the top 5 pathway of KEGG and three differential genes, with different colored lines corresponding to different KEGG pathways.

Fig 3. Expression and correlation of the hub DEGs.

(A) The volcano map of the seven differential genes presented separately. (B) Expression analysis of 7 differential genes in UC and healthy controls (ggplo2 package mapping). ns p>0.05, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. (C) 7 DEG-encoded protein interaction networks. The network nodes represent proteins, while the lines indicate predicted relationships: with light blue representing auxiliary database evidence, purple representing laboratory proof, yellow representing text mining evidence, green representing gene similarity, red representing gene fusion, blue representing gene co-production, black lines representing gene co-expression, and gray lines representing protein homology.

Fig 4. Machine learning screening for key differential genes.

(A) LASSO regression screening of 5 genes. (B) RF selected 5 genes in order of importance. (C) SVM screened 6 genes. (D) Intersection obtained 4 core genes. (E) The correlation between the 4 core genes, with red represents a positive correlation, and green indicating a negative correlation. (F) ROC curve of 4 genes predicting disease occurrence.

Fig 5. Evaluation of the degree of immune cell infiltration.

(A) Correlation analysis between immune cells. (B) Differences in immune cell infiltration between UC and healthy control group, ns p>0.05, *p<0.05, **p<0.01, ***p<0.001. (C) Correlation analysis between 4 core genes and immune cells.

Fig 6. Single gene enrichment analysis.

(A) GSEA analysis for ADAMTS1. (B) GSEA analysis for GXYLT2. (C) GSEA analysis for MUC1. (D) GSEA analysis for SEMA5A.

Fig 7. Identification and validation of ulcerative colitis subtypes.

(A) Heatmap of sample clustering at consensus k = 2. (B) The expression status of four hub genes in the two subtypes, ***p<0.001. (C) Heatmap of four hub genes in the two subtypes.

Fig 8. The diversity of the underlying biological function characteristics between the two subtypes.

(A) The differences in KEGG pathway enrichment score between subtypes A and B. (B) The differences in Reactome pathway enrichment score between subtypes A and B.

Fig 9. Differential genes and enrichment analysis of the two subtypes.

(A) PCA analysis demonstrating a distinctive difference between the two clusters. (B) Volcano plot of the 229 DEGs. The threshold for the volcano plot was |logFC| >0.5 and adj.p.Val. < 0.05. (C) GO enrichment analysis showing the BP, CC, and MF parts. (D) The bubble plot depicting the KEGG pathway enrichment analysis of DEGs. (E) The correspondence between the KEGG top five pathways and genes.

Fig 10. TF–miRNA co-regulatory network analysis, with red nodes representing hub genes, and blue nodes indicating TFs and miRNAs.