Identification of immune cell infiltration and effective biomarkers of polycystic ovary syndrome by bioinformatics analysis

Abstract

Background: Patients with polycystic ovary syndrome (PCOS) exhibit a chronic inflammatory state, which is often accompanied by immune, endocrine, and metabolic disorders. Clarification of the pathogenesis of PCOS and exploration of specific biomarkers from the perspective of immunology by evaluating the local infiltration of immune cells in the follicular microenvironment may provide critical insights into disease pathogenesis.

Methods: In this study, we evaluated immune cell subsets and gene expression in patients with PCOS using data from the Gene Expression Omnibus database and single-sample gene set enrichment analysis.

Results: In total, 325 differentially expressed genes were identified, among which TMEM54 and PLCG2 (area under the curve = 0.922) were identified as PCOS biomarkers. Immune cell infiltration analysis showed that central memory CD4⁺ T cells, central memory CD8⁺ T cells, effector memory CD4⁺ T cells, γδ T cells, and type 17 T helper cells may affect the occurrence of PCOS. In addition, PLCG2 was highly correlated with γδ T cells and central memory CD4⁺ T cells.

Conclusions: Overall, TMEM54 and PLCG2 were identified as potential PCOS biomarkers by bioinformatics analysis. These findings established a basis for further exploration of the immunological mechanisms of PCOS and the identification of therapeutic targets.

Keywords: Biomarker; Immune cell infiltration; Polycystic ovary syndrome; Single-sample gene set enrichment analysis.

Fig. 1

Two-dimensional PCA and boxplot before and after removing interbatch effects. (A, B) Data before removing interbatch effects in datasets GSE84958, GSE106724, GSE114419, and GSE137684. (C, D) Data after removing interbatch effects

Fig. 2

Differences in immune cell infiltration between patients with PCOS and normal controls. (A) Differences in the enrichment of immune cells between samples from the two groups. The normalised relative abundances of immune cells in each sample are shown, with red representing high enrichment degree and green representing low enrichment degree. (B) Correlations of 28 types of immune cells in the dataset. Colour blocks with a correlation coefficient p value greater than 0.05 in the upper part of the graph are not displayed; red represents a positive correlation, and green represents a negative correlation, with darker colour indicating a stronger correlation. (C) Differences in the enrichment of 28 types of immune cells between patients with PCOS and normal controls. Results with p values less than 0.05 are indicated in red, and immune cells with different enrichment degrees on the axis are indicated as red or blue. Red indicates a significant increase in the PCOS group, whereas blue indicates a significant decrease in the PCOS group

Fig. 3

Analysis of DEGs. (A) Volcano diagram of DEGs between patients with PCOS and normal controls. Green represents significantly downregulated genes in the PCOS group, red represents significantly upregulated genes in the PCOS group, and black represents genes with no difference between groups. (B) Heat map of the top 50 genes with larger absolute value differences in expression between sample groups. The colour of the block represents the normalised expression of corresponding genes in the sample

Fig. 4

Functional correlation analysis of DEGs. (A) Top 20 results from GO enrichment analysis of DEGs between the PCOS and normal control groups. The shape of the pattern indicates the annotation of GO, the size of the pattern indicates the number of enriched genes, and the colour of the pattern indicates the p value. (B) Top 20 results from KEGG enrichment analysis of DEGs between the PCOS and normal control groups. The size of the pattern indicates the number of enriched genes, and the colour of the pattern indicates the p value. (C) Partial results from DO analysis of DEGs between the PCOS and normal control groups. The abscissa indicates the number of enriched genes, and the colour indicates the p value

Fig. 5

Screening of PCOS biomarkers. (A, B) LASSO regression analysis was used to screen for biomarkers. Different coloured lines represent different genes. (C) BORUTA algorithm for feature variable screening. (D) The intersection results of the two algorithms were obtained using a Venn diagram to yield candidate biomarkers

Fig. 6

Expression of candidate biomarkers. (A) Heatmap of differences in the expression of candidate biomarkers between PCOS and normal control samples. The colours of the blocks represent the normalised gene expression levels in the samples. (B) Boxplot of differences in the expression of candidate biomarkers between groups. *** p < 0.001, and **** p < 0.0001. (C) Heatmap of the correlations between 20 candidate biomarkers. Colour blocks with correlation coefficients greater than 0.05 in the upper part of the figure are not displayed. Blue represents positive correlations, orange represents negative correlations, and darker colours indicate stronger correlations

Fig. 7

Correlations between candidate biomarkers and 28 types of immune cells. (A) Heatmap of the correlations between 28 immune cells and 20 candidate biomarkers. The colours of the blocks indicate the magnitude of the correlation. (B, C) Scatter plots of the two groups of correlations with the largest absolute values of positive and negative correlations

Fig. 8

Predictive value of biomarkers. (A) The predictive value of 20 candidate genes for PCOS was evaluated using a bar graph of the AUC. An AUC value greater than 0.85 indicated that the model differentiating effect was satisfactory. (B) Diagnostic efficacy comparison of the ROC curves of TMEM54 and PLCG2 combined and separate

Fig. 9

(A) Diagnostic efficacy of the ROC curves of TMEM54 and PLCG2 combined in RT-qPCR. (B) Diagnostic efficacy of the ROC curves of TMEM54 and PLCG2 combined in GSE193812.