Identification and validation of potential diagnostic signature and immune cell infiltration for NAFLD based on cuproptosis-related genes by bioinformatics analysis and machine learning

Figures

• 

  Figure 1. Flowchart of the present study.

• 

  Figure 2. Identification of cuproptosis-related differentially. Expressed genes in NAFLD. (A) Venn diagram showed the intersection of genes between DEGs and cuproptosis-related genes. (B) The locations of the 9 DE-CRGs on 23 chromosomes. (C) Boxplots showed the differential expression of CRGs between NAFLD and control samples. (D) The expression patterns of 9 DE-CRGs were shown in the heatmap. (E) Correlation analysis of 9 DE-CRGs. Red and Blue colors represent positive and negative correlations, respectively. (F) Gene relationship network diagram of 9 DE-CRGs. p values were showed as: *, p < 0.05; **, p < 0.01; ***, p < 0.001. DEG, differential expression genes; DE-CRGs, differentially expressed cuproptosis-related genes.

• 

  Figure 3. Functional analyses for the DE‐CRGs. (A) Bubble diagrams of the GO enrichment analysis of 9 DE-CRGs. (B) Bubble diagrams of the KEGG enrichment analysis of 9 DE-CRGs. DE-CRGs, differentially expressed cuproptosis-related genes. GO, Gene Ontology; BP, Biological process; CC, Cellular component; MF, Molecular function; KEGG, Kyoto Encyclopedia of Genes and Genomes.

• 

  Figure 4. Identification of diagnostic marker genes for NAFLD through machine learning. (A) Ten-time cross-verification of adjusted parameter selection in the LASSO model. Each curve corresponds to one gene. (B) LASSO coefficient analysis. The solid vertical lines represent the partial likelihood deviance SE. The dotted vertical line is drawn at the optimal lambda. (C) Relationship between the number of random forest trees and error rates. The red line represents the error of the NAFLD group, the green line represents the error of the Control group, and the black line represents the total sample error. (D) The rank of genes in accordance with their relative importance. (E) The accuracy and (F) the error of the feature selection for the SVM-RFE algorithm. The peak of the curve is achieved at 3 genes with an accuracy of 63%, with the lowest cross-validation error is found in 3 gens and the values is 36.2%.(G) The Venn diagram shows the overlap of marker genes between LASSO, random forest, and SVM-RFE algorithms.

• 

  Figure 5. Validation of marker gene expression. (A) Logistic regression model to identify the AUC of NAFLD. The AUC value was 0.704(95%CI 0.632-0.774). (B) The ROC results for the 3 marker genes. The AUC value of DLD, POLD1, and NFE2L2 was 0.58, 0.616, 0.639, respectively. (C) Nomogram graph of the 3 marker genes. (D) Calibration curve displaying the diagnostic ability of the nomogram model. (E) DCA illustrating the predictive efficiency of Nomogram models. (F) The clinical impact curve showed a higher diagnostic ability of the nomogram model. (G) Boxplots indicating the three differentially expressed CRGs were significant alter between NAFLD and control samples in GSE135251. (H) The ROC results of 3 marker genes in GSE135251. The AUC value of DLD, POLD1, and NFE2L2 was 0.729, 0.963, 0.774, respectively. (I) ROC results of the 3-gene-based model based on 3-fold cross-validation in GSE135251. The AUC value as 0.970. AUC, area under curve; ROC, receiver operating characteristic; DCA, Decision curve analysis.

• 

  Figure 6. GSEA and GSVA analysis of three marker genes. The KEGG pathway enrichment analysis of (A) DLD, (B) NFE2L2 and (C) POLD1 were carried out by GSEA enrichment method, and the two items with the highest and lowest enrichment scores are visualized according to the arrangement of enrichment scores. The KEGG pathway enrichment analysis of (D) DLD, (E) NFE2L2 and (F) POLD1 was carried out by GSVA enrichment method, and the top 50 are visualized according to the enrichment score.

• 

  Figure 7. Immune Infiltration analysis between NAFLD and controls by CIBERSORT and ssGSEA algorithm. (A) The relative abundances of 22 infiltrated immune cells between NAFLD and control samples. (B) Boxplots indicated the differences in immune infiltrating between NAFLD and control samples. The results showed 9/22 of immune cells were significantly different between NAFLD and control group. (C) The correlation between 22 immune cells and three marker genes. Red and green colors represent positive and negative correlations, respectively. (D, E) Boxplots indicated the differences in immune cells and function between NAFLD and control samples. (F) The correlation between 29 immune cells and functions and three marker genes. Red and green colors represent positive and negative correlations, respectively. p values were showed as: *, p < 0.05; **, p < 0.01; ***, p < 0.001.

• 

  Figure 8. Gene-targeted drugs and ceRNA networks. (A) mRNA-drugs interaction network. The green circular node represented the drugs. (B) The ceRNA network based on marker genes. With Pink dots for mRNA, orange dots for miRNA, and blue dots for lncRNA.

• 

  Figure 9. NAFLD mice model construction and Altered expression of CRGs in NAFLD. (A, B) serum alanine aminotransferase (ALT) and aspartate aminotransferase (AST) value in both wildtype and NAFLD mice. (C) The HE and Oil Red staining in wildtype and NAFLD mice. The result of HE and Oil red staining showed numerous fat droplets deposition in NAFLD group. (D-F) mRNA expression of DLD, POLD1, NFE2L2 by RT-PCR. Liver tissues and blood of the wildtype and NAFLD mice were collected. p values were showed as: *, p < 0.05; **, p < 0.01; ***, p < 0.001. CRGs, cuproptosis-related genes. Wild-type control mice(n=8), NAFLD mice(n=8).