Identification and validation of oxidative stress-related genes for the diagnosis of sepsis-induced acute lung injury

Abstract

Sepsis-induced acute lung injury (ALI) is an inflammatory pulmonary condition characterized by a complex pathophysiological mechanism. The development and progression of sepsis-induced ALI are accompanied by significant oxidative damage. This study aimed to identify key oxidative stress-related genes associated with sepsis-induced ALI. Samples, including sepsis, sepsis-induced ALI, and control groups, were obtained from the Gene Expression Omnibus database. Key oxidative stress-related genes in sepsis-induced ALI were identified using Weighted Gene Co-expression Network Analysis (WGCNA), Protein-Protein Interaction (PPI) network analysis, logistic regression, and LASSO regression analysis. Functional information regarding these genes was explored through Gene Set Variation Analysis (GSVA) and Gene Set Enrichment Analysis (GSEA). A logistic regression model was constructed based on the identified hub oxidative stress-related genes. The diagnostic value of this model for sepsis-induced ALI was assessed using the receiver operating characteristic (ROC) curve. The relative abundance of 22 human immune cell types was calculated using CIBERSORT software. The expression levels of hub genes in the blood samples of sepsis-induced ALI patients were analyzed through RT-PCR and ELISA. A total of 1,055 genes associated with sepsis-induced ALI were identified via WGCNA, of which 145 genes were linked to oxidative stress. GSVA revealed that these 145 genes were significantly enriched in 79 biological pathways, while GSEA indicated a strong association with immune-related signaling pathways. Additionally, the top 20 genes were selected through PPI network analysis. The logistic regression model was constructed using VDAC1, HSPA8, SOD1, HSPA9, TXN, and SNCA. In the training set and the validation set, the AUC values of logistic regression model were 0.9091 and 0.8279, respectively, suggesting good discriminability when distinguishing normal from sepsis-induced ALI. Notably, these six genes were correlated with immune cell infiltration in sepsis-induced ALI, with HSPA8, SOD1, and HSPA9 showing downregulation in sepsis-induced ALI. In conclusion, VDAC1, HSPA8, SOD1, HSPA9, TXN, and SNCA have been identified as oxidative stress-related genes associated with sepsis-induced ALI. The logistic regression model developed using these six genes could identify patients with sepsis-induced ALI. Our findings might provide novel research strategies for the molecular therapeutic target of sepsis-induced ALI.

Fig 1. The workflow of this article.

Fig 2. Identification of sepsis-induced ALI onset related genes.

A. The heat map of module genes. B. The expression level of genes in each module. C. The correlation of gene modules with sepsis-induced ALI. The color bar on the far right illustrates the range of correlation. Red indicates a positive correlation, and blue signifies a negative correlation. Darker shades indicating stronger correlations. Each number in the cells denotes the correlation value and its significance.

Fig 3. Identification of oxidative stress-related target genes in sepsis-induced ALI.

A. The overlapping genes between sepsis-induced ALI-related genes group and oxidative stress-related genes group. B. The result of the PPI network. The enriched pathways of 145 sepsis onset-related genes using GSVA (C) and GSEA (D).

Fig 4. Diagnostic model construction for sepsis-induced ALI.

A. The top 20 genes in the PPI network. B. Normality of residuals was confirmed through QQ plot of residuals. The normal QQ plot can be used to test whether a data sequence conforms to a certain probability distribution. It creates a scatter plot with the probability quantiles of the corresponding distribution as the horizontal axis and the quantiles of the data sequence as the vertical axis. C. The component residual plots for the 6 genes in the logistic regression model. The presence of a noticeable linear relationship between the horizontal and vertical axes in the graph suggests that the independent variable is appropriate for inclusion in the model. D. The leverage plot of residual. The ROC curve of the model in meta-GEO cohort (E), validation set (F), and GSE66890 dataset (G).

Fig 5. Top 7 significantly enriched Reactome pathway.

Fig 6. The regulatory network of TF-miRNA-mRNA.

Red diamonds represent model genes, blue triangles represent miRNAs, and yellow circles represent TFs.

Fig 7. Correlation of VDAC1, HSPA8, SOD1, HSPA9, TXN, and SNCA with immune cell infiltration in sepsis-induced ALI.

(A) The relative contents of 22 immune cell infiltration of samples in the meta-GEO cohort. The proportion of immune cell infiltration in the sepsis-induced ALI and control samples in the meta-GEO cohort calculated by CIBERSORT (B) and XCELL algorithms (C). The correlation between the proportion of immune cell infiltration and VDAC1, HSPA8, SOD1, HSPA9, TXN and SNCA expression calculated by CIBERSORT (D) and XCELL algorithms (E). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Fig 8. The expression of hub gene expression in sepsis-induced ALI.

A. The expression of VDAC1, HSPA8, SOD1, HSPA9 in sepsis-induced ALI in the meta-GEO cohort. B. The level of HSPA8 and HSPA9 miRNA expression in the blood samples of patients with sepsis-induced ALI. C. The concentration of HSPA8 and HSPA9 protein in the serum of patients with sepsis-induced ALI. *p < 0.05, **p < 0.01.