Novel anoikis-related genes for the diagnosis of non-obstructive azoospermia

Abstract

Background: Non-obstructive azoospermia (NOA) is a prevalent cause of male infertility, featured by the absence of sperm in the ejaculate due to impaired spermatogenesis. The involvement of anoikis in the pathogenesis of NOA remains inadequately understood. This research aims to identify anoikis-related genes as potential biomarkers for NOA diagnosis.

Methods: Based on the Gene Expression Omnibus (GEO) database and Limma R package, we identified differentially expressed genes of NOA and downloaded anoikis-related genes based on the GeneCards database. Subsequently, anoikis-related hub genes were screened by machine learning (ML), and validated using external validation sets. A nomogram constructed from these genes demonstrated high predictive accuracy, while boxplots and complex heatmaps illustrated the differential expression patterns observed in NOA samples. Additionally, immune infiltration analysis was performed using the CIBERSORT algorithm to evaluate the distribution of immune cells in both NOA and control groups. The validation of candidate genes was conducted through receiver operating characteristic (ROC) curve analysis, with the area under the curve (AUC) indicating predictive accuracy.

Results: Ultimately, we screened three hub genes: GLO1, BAP1, and PLK1. GLO1 was found to be up-regulated, while both BAP1 and PLK1 were down-regulated. Immune cell infiltration analysis elicited significant differences in 16 immune cell types between NOA patients and normal controls, with Tregs and macrophages notably up-regulated in NOA. ROC analysis indicated that all the three hub genes exhibited excellent diagnostic efficacy. Specifically, ROC curve analysis confirmed the diagnostic potential of GLO1, BAP1, and PLK1, yielding AUC values of 0.981, 0.980, and 0.981 in internal datasets, and 0.750, 0.875, and 1.000 in external datasets.

Conclusions: By ML analysis, this research identified three anoikis-related genes that may be diagnostic biomarkers for NOA, offering views into the underlying molecular mechanisms and therapeutic targets.

Keywords: Anoikis-related genes; non-obstructive azoospermia (NOA); random forest (RF); support vector machine (SVM).

Figure 1

Flowchart of this study. DEG, differentially expressed gene; GO, gene ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; NOA, non-obstructive azoospermia; ROC, receiver operating characteristic; SVA, Surrogate Variable Analysis.

Figure 2

DEGs in NOA and normal samples. (A) Volcano diagram of DEGs. Red dots signify the significantly downregulated genes, and blue dots signify up-regulated genes. (B) Venn diagram of the 45 anoikis-related DEGs. DEGs, differentially expressed genes; FC, fold change; NOA, non-obstructive azoospermia.

Figure 3

GO and KEGG functional enrichment analyses. (A) BP that DEGs enriched in. (B) CC that DEGs enriched in. (C) MF that DEGs enriched in. (D) KEGG enrichment results. BP, biological process; CC, cellular component; DEGs, differentially expressed genes; GO, gene ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; MF, molecular function.

Figure 4

Three candidate genes were screened using RF and SVM. (A) Error rates were plotted using the randomForest package, and the highest diagnostic accuracy was found at the 66th combination. (B) RF-RFE uncovered nine key genes. (C) The screened key genes were sorted by importance. (D) Twelve key genes were found using SVM. (E) Candidate genes obtained from RF and SVM were intersected to obtain 3 candidate genes. RF, random forest; RF-RFE, random forest-recursive feature elimination; SVM, support vector machine.

Figure 5

Construction of the nomogram, Boxplot, and complex heatmap of three key genes. (A) The visible nomogram for diagnosing NOA. (B) Complex heatmaps of the expression of 3 genes in the NOA and normal groups; (C) Boxplots of the expression of 3 candidate genes in the NOA and normal groups. ****, P<0.0001. NOA, non-obstructive azoospermia.

Figure 6

Immune infiltration analysis. Sixteen immune cells showed significant differences in infiltration between NOA patients and normal controls. *, P<0.05; **, P<0.01; ***, P<0.001; ****, P<0.0001; ns, not significant. NOA, non-obstructive azoospermia.

Figure 7

Diagnostic efficacy and AUC values of 3 key genes in the training set and external validation dataset. The ROC curves of the diagnostic model for 3 key genes in the training set (A-C). The ROC curves of the diagnostic model for 3 key genes in the validation set (D-F). AUC, area under the curve; ROC, receiver operating characteristic.