Exploring the pathogenesis, biomarkers, and potential drugs for type 2 diabetes mellitus and acute pancreatitis through a comprehensive bioinformatic analysis

Abstract

Background: Type 2 diabetes mellitus (T2DM) is a chronic metabolic disease that accounts for > 90% of all diabetes cases. Acute pancreatitis (AP) can be triggered by various factors and is a potentially life-threatening condition. Although T2DM has been shown to have a close relationship with AP, the common mechanisms underlying the two conditions remain unclear.

Methods: We identified common differentially expressed genes (DEGs) in T2DM and AP and used functional enrichment analysis and Mendelian randomization to understand the underlying mechanisms. Subsequently, we used several machine learning algorithms to identify candidate biomarkers and construct a diagnostic nomogram for T2DM and AP. The diagnostic performance of the model was evaluated using ROC, calibration, and DCA curves. Furthermore, we investigated the potential roles of core genes in T2DM and AP using GSEA, xCell, and single-cell atlas and by constructing a ceRNA network. Finally, we identified potential small-molecule compounds with therapeutic effects on T2DM and AP using the CMap database and molecular docking.

Results: A total of 26 DEGs, with 14 upregulated and 12 downregulated genes, were common between T2DM and AP. According to functional and DisGeNET enrichment analysis, these DEGs were mainly enriched in immune effector processes, blood vessel development, dyslipidemia, and hyperlipidemia. Mendelian randomization analyses further suggested that lipids may be a potential link between AP and T2DM. Machine learning algorithms revealed ARHGEF9 and SLPI as common genes associated with the two diseases. ROC, calibration, and DCA curves showed that the two-gene model had good diagnostic efficacy. Additionally, the two genes were found to be closely associated with immune cell infiltration. Finally, imatinib was identified as a potential compound for the treatment of T2DM and AP.

Conclusion: This study suggests that abnormal lipid metabolism is a potential crosstalk mechanism between T2DM and AP. In addition, we established a two-gene model for the clinical diagnosis of T2DM and AP and identified imatinib as a potential therapeutic agent for both diseases.

Keywords: acute pancreatitis; biomarker; machine learning; molecular docking; type 2 diabetes mellitus.

Figure 1

The process of data analyzing in this study.

Figure 2

Identification of DEGs. (A, B) PCA analysis before and after merging of T2DM datasets; (C) The volcano plot for T2DM (|log2 FC| > 0.585 and adjust p <0.05); (D) The volcano plot for AP (|log2 FC| > 0.585 and adjust p <0.05); (E) The intersection of AP up-regulated DEGs and T2DM up-regulated DEGs; (F) The intersection of AP down-regulated DEGs and T2DM down-regulated DEGs; (G) The location of DEGs on chromosomes.

Figure 3

Functional enrichment analysis of DEGs. (A) PPI of the DEGs; (B) The GeneMANIA analysis for DEGs; (C, D) Functional and DisGeNET enrichment analyses by the Metascape database.

Figure 4

The characteristic genes of AP were screened by machine learning method. (A) LASSO regression analysis and (B) cross-validation for identifying key genes and assessing partial likelihood deviance; (C, D) Seventeen characteristic genes found by SVM-RFE; (E, F) RF ranked the importance of all genes to get 6 genes with scores for importance greater than 2; (G) The Venn diagram exhibiting the intersection of three machine learning models.

Figure 5

The characteristic genes of T2DM were screened by machine learning method. (A) LASSO regression analysis and (B) cross-validation for identifying key genes and assessing partial likelihood deviance; (C, D) Seven characteristic genes found by SVM-RFE; (E, F) RF ranked the importance of all genes to get 8 genes with scores for importance greater than 1; (G) The Venn diagram exhibiting the intersection of three machine learning models.

Figure 6

Diagnostic effect of the two-gene model on AP. Box plots showed the expression difference in (A) ARHGEF9 and (B) SLPI between AP and normal samples, ***P < 0.001; (C) ROC curve of diagnostic performance of ARHGEF9 and SLPI for AP; (D) ROC curve of the two-gene model for AP; (E) DCA curve of the model; (F) Nomogram for forecasting AP risk; (G) The calibration curve of nomogram model prediction in AP.

Figure 7

Diagnostic effect of the two-gene model on T2DM. Box plots showed the expression difference in (A) ARHGEF9 and (B) SLPI between T2DM and normal samples, ***P < 0.001; (C) ROC curve of diagnostic performance of ARHGEF9 and SLPI for T2DM; (D) ROC curve of the two-gene model for T2DM; (E) DCA curve of the model; (F) Nomogram for forecasting T2DM risk; (G) The calibration curve of nomogram model prediction in T2DM.

Figure 8

Functional enrichment and immune cell correlation analysis of characteristic genes in AP. (A-D) Enrichment biological functions and pathways of two hub genes identified by GSEA; (E) Immune cell correlation analysis of ARHGEF9 and SLPI. *P < 0.05; **P < 0.01; ***P < 0.001.

Figure 9

Functional enrichment and immune cell correlation analysis of characteristic genes in T2DM. (A-D) Enrichment biological functions and pathways of two hub genes identified by GSEA; (E) Immune cell correlation analysis of ARHGEF9 and SLPI. *P < 0.05; **P < 0.01; ***P < 0.001.

Figure 10

The ceRNA networks, single-cell maps, and immunofluorescence of ARHGEF9 and SLPI. (A) The ceRNA network of ARHGEF9; (B) The ceRNA network of SLPI; (C) The single-cell type atlases of ARHGEF9 in the pancreatic tissues; (D) The single-cell type atlases of SLPI in the pancreatic tissues; (E) The immunofluorescence of ARHGEF9 in cell line A-431, target protein (green), nucleus (blue), microtubules (red); (F) The immunofluorescence of SLPI in cell line SiHa, target protein (green), nucleus (blue), microtubules (red).