Utilizing integrated bioinformatics and machine learning approaches to elucidate biomarkers linking sepsis to fatty acid metabolism-associated genes

Figures

• 

  Fig. 1. Framework. The GSE13904 dataset was leveraged as the principal cohort, supplemented with the GSE65682 dataset for validation purposes. Through employing a meticulous matching strategy of FAMGs, differential expression analyses were conducted and a prognostic risk model was subsequently constructed. Our methodological framework revealed a distinct subset of FAMGs exhibiting prognostic relevance to sepsis and underscored their candidacy as promising biomarkers. Subsequently, comprehensive bioinformatics analyses spanning GO, KEGG, and GSEA were conducted to unravel the functional underpinnings of these identified genes. These analyses were augmented by seamless integration with diverse databases, furnishing a multi-dimensional portrayal of the implicated FAMGs within the intricate milieu of cellular processes, signaling cascades, and gene regulatory networks. Furthermore, an intricate evaluation encompassing immune cell infiltration patterns, functional alterations, and transcriptomic variations was conducted to illuminate the immunological landscape and shed insights into transcriptional dynamics.

• 

  Fig. 2. Principal component analysis. (a) Analysis of difference. (b) Analysis of correlation. Note: High expression is indicated in red, and low expression is shown in blue. The darker the color, the more enriched the results.

• 

  Fig. 3. For FAMGs, GO, and KEGG analyses were performed. (a): The GO circle illustrates the scatter map of the selected gene’s logFC. (b): The KEGG barplot and bubble illustrate the scatter map of the logFC of the indicated gene.

• 

  Fig. 4. The development of the FAMG signature. (a): Regression of the 13 sepsis-related genes using LASSO. (b): Cross-validation is used in the LASSO regression to fine-tune parameter selection. (c–d): Accuracy and error of this model. (e): Venn. (f): An AUC test was performed on the obtained 13 key genes. AUC of 13 hub genes. (g): AUC of train group. Note: It is crucial to highlight that, despite these variabilities, all genes of AUC values for these genes consistently approximate the notable benchmark of 0.7. This collective result significantly bolsters the overall credibility, precision, and robustness of our predictive model.

• 

  Fig. 5. GSEA of analysis in ACSS1 and ACSS2. (a): GO results of ACSS1 and ACSS2. (b): KEGG results of ACSS1 and ACSS2. Note: High expression is indicated in red, and low expression is shown in blue. The longer the color, the more enriched the results.

• 

  Fig. 6. Expression of Immune cells. (a) Expression of immune cells in different clusters. (b) Correlation between FAMGs and immune cells.

• 

  Fig. 7. GSVA of Analysis in ACSS1 and ACSS2. (a): GO analysis results of ACSS1 and ACSS2. (b): KEGG analysis results of ACSS1 and ACSS2. Note: High expression is indicated in red, and low expression is shown in blue. The longer the color, the more enriched the results.

• 

  Fig. 8. Drug-gene interactions. Note: Red circles represent upregulated genes; green hexagons represent downregulated genes; and blue squares represent the associated drugs.

• 

  Fig. 9. miRNAs-LncRNAs shared Genes Network. Note: Red circles represent mRNAs; blue quadrangles represent miRNAs; and green triangles represent lncRNAs.

• 

  Fig. 10.. 13 hub genes were validated.

• 

  Fig. 11. Model verification. (a–b) Residual expression patterns. (c) Model expression patterns (d) AUC of the model. (e) AUC of the test group.