Fig 1.
Sankey Diagram in Co-expression Analysis.
Through co-expression analysis of disulfidptosis-related genes and pulmonary arterial hypertension (PAH) gene expression profiles, we identified genes associated with disulfidptosis. A total of 5,656 genes were screened, including:1,055 genes co-expressed with the disulfidptosis gene GYS1;1,105 genes co-expressed with LRPPRC;1,360 genes co-expressed with NCKAP1;1,577 genes co-expressed with NDUFA11;29 genes co-expressed with NDUFS1;158 genes co-expressed with NUBPL;8 genes co-expressed with OXSM;122 genes co-expressed with RPN1;155 genes co-expressed with SLC3A2;87 genes co-expressed with SLC7A11.
Fig 2.
Weighted Gene Co-expression Network Analysis (WGCNA) and Hub Genes Identification.
(A)shows “Sample Dendrogram and Trait Heatmap”. This figure illustrates sample similarity and phenotypic classification. The dendrogram (generated via hierarchical clustering of gene expression data) reveals relationships among samples, with shorter branches indicating higher expression similarity. The trait heatmap uses color-coding: blue for the Control group and red for the Pulmonary Arterial Hypertension (PAH/Treat) group. (B)shows “Scale Independence Plot and Mean Connectivity Plot”. The left panel displays the scale-free topology model fit (R²) across different soft-thresholding powers, while the right panel shows mean connectivity. The red line (R² = 0.89) indicates that a soft-thresholding power of 18 was selected as optimal for constructing a biologically relevant co-expression network. (C)shows “Heat Map of Module-Trait Relationships”. This heatmap depicts correlations between module eigengenes (y-axis: MEgreen, MEyellow, MEgrey) and phenotypic traits (x-axis: Control vs. PAH). Color intensity reflects correlation strength—dark yellow for strong positive, dark blue for strong negative—with p-values indicating significance. (D)shows “Gene Significance”. The histogram presents the distribution of gene significance values (x-axis) across genes (y-axis). The green module shows the highest significance, suggesting its strongest association with PAH. (G)shows “Gene Significance Scatter Plot of the Green Module”. This plot examines the relationship between module membership (x-axis) and gene significance (y-axis) in the green module, highlighting genes most strongly linked to PAH while maintaining high intramodular connectivity.
Fig 3.
Fig 3 presents the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses performed on the core genes identified in the green module. (A) shows the GO enrichment analysis, covering BP, CC, and MF. The bubble plot displays the top five significantly enriched terms, where the bubble size corresponds to the number of genes involved (larger bubbles represent higher gene counts) and the color gradient reflects the adjusted p-value significance (more intense red indicates greater statistical significance). (B) shows the KEGG pathway enrichment analysis, featuring the top 20 enriched pathways in a similar bubble plot format. The visualization highlights the most relevant metabolic and signaling pathways, with bubble size and color coding maintaining the same representation as in panel A for consistent interpretation.
Fig 4.
(A) and (B) show that the Least Absolute Shrinkage and Selection Operator(LASSO) regression algorithm was applied for feature gene selection, with the regularization parameter λ used for covariate selection and dimensionality reduction. (C) and (D) show that feature genes were screened using the Support Vector Machine-Recursive Feature Elimination(SVM-RFE) algorithm, an iterative approach that ranks and eliminates the least important features based on classifier performance. (E) and (F) show that the Random Forest(RF) algorithm was employed for feature gene selection, leveraging its built-in feature importance scoring to identify the most relevant genes. (G) shows a Venn diagram illustrating the overlapping key feature genes identified by LASSO, SVM-RFE, and RF, highlighting consensus biomarkers across different selection methods. The genes selected by Lasso include AKR7A2P1, AKR7A3, ATG3, RANBP6, TRAPPC9, TTLL1, USP32, and ZNF655. The genes selected by the SVM-RFE algorithm include USP32, ZNF655, AHR, PNRC2, ERBB2IP, ZNF687, ZC3H15, PSMD12, YTHDF3, CPNE8, and ALKBH4. The genes selected by the RF algorithm include ZNF655, AKR7A2P1, USP32, YTHDF3, and PNRC2.
Fig 5.
Area Under the Receiver Operating Characteristic Curve(AUC-ROC) Results.
(A) shows the ROC curve of gene USP32 in the training set GSE15197 (GPL6480), demonstrating its diagnostic performance. (B) shows the ROC curve of gene USP32 in the independent validation set GSE113439 (GPL6244), confirming its robustness. (C) shows the ROC curve of gene ZNF655 in the training set GSE15197 (GPL6480), evaluating its classification efficacy. (D) shows the ROC curve of gene ZNF655 in the validation set GSE113439 (GPL6244), further validating its predictive capability.
Fig 6.
Fig 6 shows the GSEA-predicted potential pathways and mechanistic functions of hub genes in the GSE15197 dataset. (A)and (B) show GO and KEGG enrichment analyses for the USP32 high-expression group, revealing its potential biological roles and associated pathways. (C) and (D) show GO and KEGG enrichment analyses for the USP32 low-expression group, highlighting distinct functional mechanisms compared to the high-expression group. (E) and (F) show GO and KEGG enrichment analyses for the ZNF655 high-expression group, identifying key biological processes and signaling pathways linked to its overexpression. (G) and (H) show GO and KEGG enrichment analyses for the ZNF655 low-expression group, uncovering differential pathway activities relative to high-expression conditions.
Fig 7.
The Violin Plot Obtained From Immune Infiltration Analysis.
(A) shows a heatmap depicting the differential distribution of 28 immune cell subtypes across study samples, highlighting distinct immune infiltration patterns between the PAH group and controls. (B) presents violin plots comparing the infiltration levels of 28 immune cell populations between control subjects and PAH patients, demonstrating significant differences in immune cell abundance between groups. (C) displays box plots illustrating the expression differences of hub genes (USP32 and ZNF655) between control and PAH groups, revealing their potential roles in disease pathogenesis.
Fig 8.
ceRNA regulatory network diagram——USP32.
Fig 8 shows the ceRNA regulatory network diagram of miRNA and lncRNA related to USP32.
Fig 9.
ceRNA regulatory network diagram——ZNF655.
Fig 9 shows the ceRNA regulatory network diagram of miRNA and lncRNA related to ZNF655.