Exploring candidate biomarkers for nonobstructive azoospermia: A bioinformatics and machine learning approach.

ObjectiveNonobstructive azoospermia (NOA), a severe male reproductive disorder, has a complex pathogenesis. This study aimed to identify novel oxidative stress potential biomarkers for NOA using bioinformatics.MethodsDatasets GSE9210, GSE108886, and GSE45885 from the GEO database were used for training, while GSE145467 supported weighted gene coexpression network analysis. The limma package identified differentially expressed genes, cross-referenced with oxidative stress genes from GeneCards for drug prediction. Three machine learning algorithms identified potential hub genes, validated by GSE45887 and GSE216907. Advanced analyses focused on these hub genes, including regulatory networks and single-gene enrichment studies. qRT-PCR validated hub gene expression, and single-cell RNA sequencing (GSE149512 and GSE202647) explored cellular expression patterns.ResultsIL1RN, NOSIP, PDYN, PRKCZ, and UTRN were potential NOA markers. qRT-PCR validation confirmed IL1RN, NOSIP, PRKCZ, and UTRN align with bioinformatic results. Single-cell analysis showed stage-specific dysregulation of these four genes in NOA.ConclusionsThis study broadens the spectrum of known NOA candidate biomarkers and enhances the comprehension of NOA.