Protein interactions, network pharmacology, and machine learning work together to predict genes linked to mitochondrial dysfunction in hypertrophic cardiomyopathy

Figures

• 

  Fig. 1. The technical route of this article.

• 

  Fig. 2. Differential Expression Analysis and weighted gene co-expression network analysis (WGCNA) of different expression genes (DEGs). (A) Based on differential expression analysis, a volcano map displays down-regulated genes (blue), no significant genes (grey), and up-regulated genes (red). (B) A heatmap that clusters the DEGs of two groups and displays how their expressions differ from one another. (C) Gathering samples and organizing them into several clinical categories. (D) The relationship between the scale-free fit index and various soft-thresholding powers. (E) The relationship between the mean connection and various soft-thresholding capabilities. (F) Unique modules are indicated by different colors in the gene clustering tree diagram. (G) Assembling clusters according to different values of module features. (H) The clinical characteristics and the relationships among the 14 modules. (I) The scatterplot in the blue module that shows how membership (MM) and gene significance (GS) are related.

• 

  Fig. 3. Functional enrichment analysis. (A) The Gene Ontology (GO) enrichment results for the blue module, broken down into molecular functions (MF), cellular components (CC), and biological processes (BP). (B, C) The blue module illustrates the enrichment of the Kyoto Encyclopedia of Genomes (KEGG) pathway. (D) Samples with Gene Set Variation Analysis (GSVA) are displayed.

• 

  Fig. 4. The immune microenvironment of HCM is the focus of this study. (A–C) We utilized an xCell analysis box plot to display the immunization ratings of the aggregated samples. We used ESTIMATE analysis to find the stromal score, the immune score, and the estimate score for HCM samples. (E) We analyzed the samples for HCM using the EPIC immunoassay. (F) MCPcounter research revealed that 10 immune-related cells were increased in the sample. (G) A box plot from the CIBERSORTX analysis shows that the sample had an enrichment of 22 immune-related cells. (H–K) The heat map and box plot were produced by the ssGSEA analysis of the samples, respectively. We used the rank sum test to analyze the data. The data were shown as the mean ± SD for all n = 3. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001 versus the control group.

• 

  Fig. 5. Protein–protein interaction (PPI) networks and machine learning work together to screen for the most characteristic gene (MCG). (A, B) Feature gene screening with the Lasso model. We identify non-zero coefficients in the optimal model by utilizing rainbow and binomial deviation plots. (C, D) IncMSE and IncNodePurity Random Forest Analysis is used to display the top variables. (E) A branch diagram showing decision trees filtering for characteristic variables. (F, G) The picture shows the features of the ten genes by using the support vector machine-recursive feature elimination (SVM-RFE) method. We can find the optimal n = 6 for training RMSE by varying the quantity of variables. (H, I) Force and waterfall diagrams are used to display the SHapley Additive exPlanations (SHAP) values of the blue module genes, which were determined using the XGBoost model. (J) Waterfall charts illustrate the blue module genes’ SHAP values, which were determined using the LightGBM model. (K, L) Bar graphs displaying the most highly characterized variables in the XGBoost and LightGBM models, respectively. (M) For visualization, blue module genes are imported into Cytoscape via the PPI network. (N)…

• 

  Fig. 6. Validation ladders of the most characteristic gene (MCG). (A) Transcriptomic data from three datasets, GSE32453, GSE35229, and GSE1145, were pooled into a validation ladder. (B) The validation ladder displays the expression levels of mRNA. (C, D) The ROC curves represent the MCG diagnostic value in the experimental and validation ladder, respectively. For the data statistics, the t-test was employed. Data were presented as the mean ± SD (n ≥ 3). *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001 versus the Sham group or Control group.

• 

  Fig. 7. Gross morphological changes in the HCM and neonatal mouse primary cardiomyocyte stimulation model. (A, B) Representative plots of mouse echocardiograms, LVEF, and LVFS statistics. (C) Typical diagrams of large mouse heart specimens. (D) Statistical Plot of HW/TL (mg/mm) and HW/BW (mg/g). (E) a-b are respectively HE staining, Masson staining, Picro-Sirius Red (PSR) stain, and WGA diagrams of representation (50um and 1000um). (F) The three statistical graphs count the area of Masson staining, Picro-Sirius Red (PSR) stain, and WGA and the size of the cells, respectively. (G, H) The Western blot demonstrated the expression of ANP and -MHC protein and mRNA levels in an animal model of HCM. (I and J) The levels of ANP and β-MHC protein and mRNA were examined in a cell model made of newborn mouse primary cardiomyocytes. For the data statistics, the t-test was employed. Data were presented as the mean ± SD (n ≥ 3). *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001 versus the Sham group or Control group.

• 

  Fig. 8. Expression of the most characteristic gene (MCG) in vivo and in vitro. (A, B) Representative graphs of TUNEL staining showing apoptosis of sections from TAC and control groups and corresponding statistics. (C, D). The mRNA level of CEBPD expression in vivo and in vitro is shown by the box plot. (E, F) The study used Western blotting to demonstrate the expression of CEBPD in animal tissues, along with the corresponding statistics. For the data statistics, the t-test was employed. Data were presented as the mean ± SD (n ≥ 3). *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001 versus the Sham group or Control group.

• 

  Fig. 9. Prediction of targeting drugs for HCM by Cmap. (A) The graph plots the RAW cs, Norm cs, doses, times, N-samples, and fdr_q_nlog10 of the Top 10 targeted drugs. (B–K) Each of the ten molecular docking diagrams represents the binding of Adarotene, Aripiprazole, Ciclesonide, Abt-751, Dexketoprofen, Medrysone, Resiquimod, Rimexolone, Tazarotene, and Tolmetin to the most characteristic gene (CEBPD). (L) The bars represent the binding strength of the ten drug monomers to CEBPD.