Identification of crucial genes for polycystic ovary syndrome and atherosclerosis through comprehensive bioinformatics analysis and machine learning

FIGURE 1. Study work flow chart. AS, atherosclerosis; DEGs, differentially expressed genes; GO, gene ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; LASSO, Least absolute shrinkage and selection operator; PCOS, polycystic ovary syndrome; RF, random forest; SVM‐RFE, support vector machine‐recursive feature elimination; WGCNA, weighted gene co‐expression network analysis.

FIGURE 2. Removal of batch effects and identification of differentially expressed genes (DEGs) in polycystic ovary syndrome (PCOS) and atherosclerosis. (a, b) Principle component analysis (PCA) plots showed the expression pattern in two data sets of PCOS before and after eliminating the batch effects. (c, d) DEG volcano and heatmap plot in PCOS group. (e, f) PCA plots showed the expression pattern in two data sets of atherosclerosis before and after removing the batch effects. (g, h) Volcano and the heatmap plot of DEGs in atherosclerosis group.

FIGURE 3. Weighted gene co‐expression network analysis of polycystic ovary syndrome (PCOS) and atherosclerosis. (a) Determination of soft‐threshold power for PCOS. (b) Cluster dendrogram of PCOS highly connected genes in key modules. (c) Relationships between modules and traits in PCOS. Correlations and p values are included in each cell. (d) Calculation of soft‐threshold power for atherosclerosis. (e) Cluster dendrogram of atherosclerosis modules with highly connected genes. (f) Module‐trait relationships in atherosclerosis. A correlation and p value are included in each cell.

FIGURE 4. Shared gene signatures and functional enrichment between polycystic ovary syndrome (PCOS) and atherosclerosis (AS). (a) Shared differentially expressed genes (DEGs) between PCOS and AS by overlapping their DEGs. (b) Shared genes between the weighted gene co‐expression network analysis (WGCNA) modules of PCOS and AS by overlapping them. (c, d) Shared genes were represented by bar plots displaying GO and KEGG enrichment. GO, gene ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes.

FIGURE 5. Shared gene protein–protein interactions network and functional enrichment of polycystic ovary syndrome and atherosclerosis. (a) Interaction network of proteins using the STRING database. The remaining 15 differentially expressed genes were eliminated due to lack of interaction. (b) Top 15 hub genes of degree by Cytohubba. (c) Gene ontology (GO) analysis (biologic process, cellular component, and molecular function) of 15 common genes. (d) Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of 15 common genes.

FIGURE 6. Candidate biomarker identification via machine learning algorithms. (a) Based on the Lasso regression algorithm, 14 genes were identified as the biomarkers with the lowest binominal deviation. (b) Thirteen genes were screened by SVM‐RFE. (c) Screening genes by importance score based on random forest algorithm. (d) Six hub genes (CD163, LAPTM5, TNFSF13B, MS4A4A, FGR, IRF1) visualize the intersection of three machine learning algorithms by Venn diagram.

FIGURE 7. Construction of the nomogram and receiver operating characteristics (ROC) curves of the six common core genes in polycystic ovary syndrome (PCOS) and atherosclerosis. (a) The ROC curve of each candidate gene (CD163, LAPTM5, TNFSF13B, MS4A4A, FGR, IRF1) was verified in the PCOS data set GSE137684. (b) The ROC curve of each candidate gene (CD163, LAPTM5, TNFSF13B, MS4A4A, FGR, IRF1) was verified in the atherosclerosis data set GSE43292. (c) Nomogram for predicting the risk of atherosclerosis in PCOS. AUC, area under the curve; CI, confidence interval.

FIGURE 8. Immune infiltration analysis between atherosclerosis and control and correlations between the hub differentially expressed genes (DEGs) and immunologic features in atherosclerosis. (a) Barplot shows the proportion of immune cells in different samples. (b) Boxplot compares the expression of immune cells between atherosclerosis and controls. *P < 0.05; **P < 0.01; ***P < 0.001. (c) Correlation analysis of immune cell infiltrations with six hub DEGs.