The workflow for this study is presented in Figure 1 . The GEOquery package in R (14) was used to extract the gene expression profile datasets GSE7014 (15) and GSE25724 (16) for T2DM from the GEO database (17). The 20 T2DM and six normal (non-diabetic) tissue samples of the GSE7014 dataset were from human (Homo sapiens) skeletal muscle biopsies, and their genomes had been sequenced using the GPL570 [HG-U133_Plus_2] Affymetrix Human Genome U133 Plus 2.0 Array platform. The six T2DM and seven normal samples of the GSE25724 dataset were human (Homo sapiens) pancreatic islet tissues, and the genes had been sequenced using the GPL96 [HG-U133A] Affymetrix Human Genome U133A Array platform. Both datasets had a common sequencing type, sample grouping information, and species source, with sufficient sample size and data quality. This was crucial for our analysis.

The limma package in R (18) was used to perform a differential analysis of the groupings on the basis of the gene expression levels in the T2DM and normal tissues. This analysis allowed for the identification of differentially expressed genes (DEGs) and their effect on the development of T2DM. First, the samples were normalized. We applied a set of filtering criteria to determine the differentially expressed PRGs. Specifically, we defined upregulated genes as those with a log fold change (FC) value of greater than 0.5 and an adjusted p-value of less than 0.05. Conversely, genes with a logFC of less than 0.5 and an adjusted p-value of less than 0.05 were classified as downregulated genes.

Additionally, we conducted a comprehensive search for pyroptosis-related studies on the PubMed database (19–33). The retrieved data were then combined with information obtained from the GeneCards database (34), AmiGO2 database (35), and Molecular Signatures Database (MSigDB) (36). This helped us identify a total of 356 PRGs ( Supplementary Table 1 ).

Next, the genes related to T2DM were obtained by intersecting the DEGs between the disease and normal groups in the two datasets. Likewise, the T2DM-PRGs were identified by examining the intersection of DEGs associated with both T2DM and PRGs.

Functional and pathway enrichment analyses of the T2DM-PRGs were conducted using Gene Ontology (GO) (37) and Kyoto Encyclopedia of Genes and Genomes (KEGG) (38) database tools, respectively. GO analysis is a common method for studying the function of genes, whereas KEGG provides information on biological pathways, diseases, and drugs. The clusterProfiler package in R (39) was used to analyze the data, with the significance threshold set at a p-value of less than 0.05.

The gene set enrichment analysis (GSEA) method was used to assess the distribution pattern of genes within a predetermined set in the gene table. This was done by ranking the genes according to their correlation with the genotype in order to determine their impact on the phenotype (40). We acquired the gene sets “c2.kegg.v7.2.symbols” and “c5.go.v7.2.symbols” from MSigDB (39) and performed GSEA on them using the clusterProfiler package in R (39), where a p-value of less than 0.05 was considered statistically significant.

The STRING database (41) is used to search for known proteins and protein–protein interactions (PPIs) and includes 2031 species, 1.38 million PPIs, and 9.6 million proteins. It contains outcomes derived from experimental data, results compiled through text mining of PubMed abstracts and other databases, and results forecasted using bioinformatic techniques. Using the STRING database, we constructed a PPI network for the DEGs associated with T2DM-PRGs.

The NetworkAnalyst database (42) is a platform used for visualizing gene expression profiles and meta-analytic data. It supports various data types from 17 species, including single or multiple gene or protein lists, single RNA sequencing or microarray gene expression data tables, multiple gene expression tables, network files, and other upload formats. In this study, we focused on analyzing the control of gene expression by miRNAs and transcription factors (TFs) at the post-transcriptional stage to identify diseases associated with the target genes (43, 44). The TarBase v8.0 (45) and ENCODE databases (46) were used to identify miRNAs and TFs associated with the differentially expressed T2DM-PRGs. We used the DrugBank v5.0 database (47) to predict the correlation between the target genes and drugs. The target DEG–miRNA, DEG–TF, and DEG–drug networks of T2DM-PRGs were visualized using Cytoscape software (48).

We analyzed the expression levels of T2DM-PRGs in both the disease and normal groups, using box plots to visualize the results. ROC curves were used to evaluate the diagnostic and predictive value of those genes, where an area under the ROC curve (AUC) value of greater than 0.7 was deemed accurate for predictive purpose. The AUC cutoff value of 0.7 is a commonly used threshold for assessing the diagnostic accuracy of a test or model. The AUC takes values from 0.5 to 1, where 0.5 indicates no discriminatory ability and 1 indicates full discriminatory ability. An AUC value of greater than 0.7 implies reasonable accuracy in diagnosing a situation.

Most tumor microenvironments consist of a combination of immune and inflammatory cells, tumor-associated fibroblasts, interstitial tissue, and various cytokines and chemokines that surround the tumor tissue. An essential part of disease research and therapy prognosis prediction is the examination of immune cell infiltration in the affected tissues.

As an extension of the GSEA method, single-sample GSEA provides the degree of enrichment of the gene set in each sample from the input data by defining the enrichment score (49). In this study, the single-sample GSEA method was used to compare immune infiltration between the diseased and normal tissues, to examine the relative increase or decrease in the occurrence of two diseases compared with the general population, and to assess the relationship between T2DM-PRGs and immune cells. Pearson’s correlation analysis was used to identify the association between the T2DM-PRGs and the level of immune invasion.

We used R v4.1.2 software for all data processing and statistical analyses. An independent Student’s t-test was applied to normally distributed variables to compare two continuous variables, whereas the Mann–Whitney U test (Wilcoxon rank sum) was used to assess differences between non-normally distributed variables. The pROC package in R was used to plot the ROC curve, and the AUC was calculated to predict patient prognosis. Statistical significance was defined as a p-value of less than 0.05.

Owing to the large number of statistical comparisons performed, multiple hypothesis testing corrections were carried out to control the probability of false discovery. Common multiple testing correction methods include the Bonferroni and Benjamini-Hochberg (also known as false discovery rate correction) procedures, which correct the original significance threshold on the basis of the sample size and desired significance level to control the overall probability of false discovery. In this study, the Bonferroni method was used for multiple testing corrections. Cross-validation and repeated experiments were used to assess the consistency and stability of the statistical model across different datasets. These help to determine whether the results are highly reliable and reduce the likelihood of false positives. Setting the appropriate significance level (e.g., 0.05 or 0.01) is an important factor in controlling the false discovery rate. Tighter significance levels will reduce the likelihood of false discoveries and lead to the omission of true differences. Independent validation was performed for the important DEGs to verify their expression or functional changes in different sample sets using other experimental methods. Additionally, biological functional analysis was conducted to further validate the biological significance of the differences, to exclude false-positive results.

First, normalization of the sample data in the two datasets was performed, and box plots of the normalized data were constructed ( Figures 2A, B ). To examine the impact of gene expression levels on T2DM tissues relative to normal tissues, the differential analysis package limma was used to generate DEGs for both datasets, which were then visualized in volcano plots ( Figures 3A, B ). In the GSE7014 dataset, 5684 DEGs were identified, 1807 of which were upregulated and 3877 were downregulated. A classification heatmap was generated ( Figure 3C ), where the DEGs were classified into two groups: diabetic and non-diabetic. In the GSE25724 dataset, 4560 DEGs were identified, 2373 of which were upregulated and 2187 were downregulated. The heatmap of the classified DEGs is shown in Figure 3D . By comparing the DEGs from both datasets, we identified 1561 genes that are associated with T2DM ( Figure 4A ). Moreover, by intersecting the PRGs with the DEGs from both datasets, we identified 25 T2DM-associated DEGs and PRGs in common ( Figure 4B ; Supplementary Table 1 ).

To examine the connections between the differentially expressed T2DM-PRGs and various biological processes, molecular functions, cellular components, biological pathways, and diseases, GO functional enrichment analysis of those genes was performed ( Figure 5A ). The differentially expressed T2DM-PRGs were mainly involved in cytokine secretion, cell junction assembly, regulation of innate immune response, positive regulation of establishment of protein localization, cell junction organization, stem cell population maintenance, calcium ion transport into the cytosol, maintenance of cell number, viral life cycle, cytosolic calcium ion transport, and other biological processes ( Figure 5B ). They were enriched in cellular components such as the cell projection membrane, COP9 signalosome, dendritic spine, neuron spine, cell leading edge, myelin sheath, sperm part, brush border membrane, sperm flagellum, and 9 + 2 motile cilium ( Figure 5C ). With regard to molecular functions, they were in high abundance for the GO terms binding to double-stranded RNA, glutamate receptors, ubiquitin protein ligases, ubiquitin-like protein ligases, and ionotropic glutamate receptors. They also showed affinity for protein phosphatase 2A and p53 ( Figure 5D ). KEGG annotation of the T2DM-PRGs revealed they were enriched in pathways related to prostate cancer, nuclear factor-kappa B (NF-κB) signaling, microRNAs in cancer, NOD-like receptor signaling, EGFR tyrosine kinase inhibitor resistance, Epstein-Barr virus infection, AGE-RAGE signaling in diabetic complications, Parkinson’s disease, thyroid hormone signaling, and autophagy. These pathways suggest the potential involvement of the T2DM-PRGs in a diverse range of diseases, including cancer, inflammation, viral infections, diabetic complications, and neurological disorders ( Figure 5E ). Specifically, the T2DM-PRGs were highly enriched in the prostate cancer (hsa05215) pathway ( Figure 5F ).

To investigate the impact of gene expression levels on T2DM, GSEA was performed on the two GEO datasets to identify the correlations between gene expression and the relevant biological processes, cellular components, and molecular functions. The results showed that in the GSE7014 dataset, the genes mainly affected adhesion molecule binding, actin cytoskeletons, muscle system processes, actin binding, ATPase activity, the regulation of actin filament-based processes, muscle tissue development, muscle organ development, muscle cell differentiation, muscle contraction, and other biological functions ( Figure 6A ). The genes in the GSE25724 dataset primarily impacted critical biological functions, such as transitions between cell cycle phases, cell division, breakdown of cellular nitrogen compounds, establishment of protein localization within organelles and the mitochondrial envelope, breakdown of modification-dependent macromolecules, restraining of the cell cycle, metabolism of nucleobase-containing small molecules, breakdown of organic cyclic compounds, and biosynthesis of organophosphates ( Figure 6B ).

With regard to the biological pathways affected by gene expression in both datasets, our findings indicated that the genes in the GSE7014 dataset primarily influenced biologically pertinent ones, including those related to focal adhesion; dilated cardiomyopathy; hypertrophic cardiomyopathy; valine, leucine, and isoleucine degradation; the citrate cycle; the TCA cycle; calcium signaling; extracellular matrix (ECM)–receptor interaction; viral myocarditis; fatty acid metabolism; and tight junctions ( Figures 6C–E ). The biological pathways that were mainly controlled by genes in the GSE25724 dataset were those related to Huntington’s disease, Alzheimer’s disease, the cell cycle, the spliceosome, ubiquitin-mediated proteolysis, oxidative phosphorylation, Parkinson’s disease, the proteasome, the citrate cycle, the TCA cycle, and protein export ( Figures 6F–H ).

In this study, the T2DM-PRGs in the STRING database were used to construct a PPI network of the differentially expressed T2DM-PRGs, using the igraph and ggraph packages in R ( Figures 7A, B ). Cytoscape software was used for visualization of the network. There were 30 T2DM-PRG-associated DEGs and 29 PPI pairs in the generated PPI network, among which DEGs related to other T2DM-PRGs interacted with one another. The five genes with the strongest cooperative relationships were PTEN, PLCG1, SIRT1, HSP90AB1, and TP63.

We constructed a T2DM-PRG–miRNA interaction network comprising 25 genes and 512 miRNAs ( Figure 8A ). The top five differentially expressed T2DM-PRGs related to prognosis were PTEN (targeted by 128 miRNAs), BRD4 (targeted by 118 miRNAs), HSP90AB1 (targeted by 103 miRNAs), VIM (targeted by 96 miRNAs), and PKN2 (targeted by 91 miRNAs).

The T2DM-PRG–TF interaction network comprised 21 genes and 274 TFs ( Figure 8B ). The top five differentially expressed T2DM-PRGs were HSP90AB1 (regulated by 151 TFs), VIM (regulated by 75 TFs), PLCG1 and SCAF11 (regulated by 54 TFs each), and PTEN (regulated by 44 TFs). The T2DM-PRG–drug interaction network included seven networks and seven genes, of which the first three networks had 92, 44, and 15 drug effects, respectively ( Figures 9A–C ).

Box plots of the expression levels of the T2DM-PRGs in the GSE7014 and GSE25724 datasets were constructed ( Figures 10A, B ). In the GSE7014 dataset, the expression levels of BCL2, CHI3L1, DDX58, DP8, PKN2, PRF1, PLCG1, PTEN, SCAF11, SDHB, SERPINB1, TP63, UBR2, and BTK were significantly different (p < 0.05). In the GSE25724 dataset, the genes with a statistically significant difference in expression level were EACE2, BRD4, BTK, CHI3L1, DRD2, DDX58, DPP8, ELAVL1, HSP90AB1, METTL3, PRF1, PLCG1, PTEN, SDHB, SEPRINB1, TP63, VIM, NLRP1, and PRKACA (p < 0.05).

In the GSE7014 dataset, the genes that exhibited diagnostic value were ACE2, BCL2, BTKCHI3L1, DDX58, DPP8, DRD2, PKN2, PLCG1, PTEN, SCAF11, SDHB, SERPINB1, and TP63 (AUC > 0.7). In the GSE25724 dataset, the genes with diagnostic value were TP63, CHI3L1, SDHB, DPP8, BCL2, TRIM31, METTL3, SEPRINB1, ACE2, DRD2, ELAVL1, UBR2, PRF1, PLCG1, PTEN, DDX58, VIM, BTK, HSP90AB1, NLRP1, and PRKACA (AUC > 0.7) ( Figure 11 ).

The permeability of T2DM and normal tissues to immune cells was compared using the single-sample GSEA method. In the GSE7014 dataset, 28 distinct types of immune cells exhibited significant differences between T2DM and normal tissue in relation to their association with T2DM-PRGs. The genes and their associated immune cells were ACE2 and gamma delta T cells; BCL2 and effector memory CD4 T cells; BRD4 and activated CD4 T cells; CHI3L1 and activated dendritic cells, central memory CD4 T cells, and central memory CD8 T cells; DPP8 and central memory CD4 T cells; DDX58 and effector memory CD4 T cells and effector memory CD8 T cells; ELAVL1 and central memory CD8 T cells; FOXO1 and central memory CD8 T cells; HSP90AB1 and central memory CD4 T cells, immature B cells, immature dendritic cells, mast cells, and memory B cells; METTL3 and central memory CD4 T cells, monocytes, natural killer cells, and neutrophils; PKN2 and central memory CD4 T cells and plasmacytoid dendritic cells; PLCG1 and activated dendritic cells; PRF1 and effector memory CD4 T cells; PTEN and central memory CD4 T cells and regulatory T cells; SCAF11 and natural killer cells; SERPINB1 and neutrophils; SDHB and central memory CD4 T cells, type 1 T helper cells, and type 2 T helper cells; TRIM31 and CD56 bright natural killer cells; TP63 and effector memory CD4 T cells; VIM and activated CD4 T cells; and UBR2 and activated CD4 T cells and type 2 T helper cells (r > 0.5, p < 0.01) ( Figure 12A ).

In the GSE7014 dataset, we observed 28 distinct types of immune cells that exhibited significant differences between the T2DM and normal tissues in relation to their association with T2DM-PRGs. The genes and immune cells that showed significant positive correlations (r > 0.5, p < 0.01) were PKN2 and activated CD4 T cells and central memory CD4 T cells; ELAVL1 and immature dendritic cells; BRD4 and activated B cells, activated dendritic cells, effector memory CD8 T cells, eosinophils, macrophages, myeloid-derived suppressor cells (MDSCs), T follicular helper cells, type 1 T helper cells, and type 2 T helper cells; UBR2 and effector memory CD4 T cells, gamma delta T cells, and immature dendritic cells; PRF1 and activated CD8 T cells, CD56 dim natural killer cells, eosinophils, and MDSCs; PTEN and activated CD4 T cells; CHI3L1 and effector memory CD4 T cells, immature dendritic cells, plasmacytoid dendritic cells, and regulatory T cells; DPP8 and effector memory CD4 T cells and immature dendritic cells; TRIM31 and activated B cells, activated dendritic cells, CD56 dim natural killer cells, eosinophils, mast cells, MDSCs, natural killer T cells, T follicular helper cells, and type 1 T helper cells; SERPINB1 and activated CD4 T cells; ACE2 and eosinophils; DRD2 and activated B cells, CD56 dim natural killer cells, eosinophils, macrophages, MDSCs, and T follicular helper cells; DDX58 and activated B cells, activated dendritic cells, macrophages, MDSCs, and neutrophils; BTK and activated B cells, eosinophils, and neutrophils; HSP90AB1 and plasmacytoid dendritic cells; NLRP1 and effector memory CD4 T cells and gamma delta T cells; and PRKACA and activated B cells ( Figure 12B ).