[Exploring the mechanisms of ferroptosis in non-obstructive azoospermia based on bioinformatics and machine learning].

Objective: To explor the potential mechanisms of ferroptosis involvement in non-obstructive azoospermia based on bioinformatics and machine learning methods.

Methods: To obtain disease-related datasets and ferroptosis-related genes, we utilized the GEO database and FerrDb database, respectively. Using the R software, the disease dataset was subjected to normalization, differential analysis, and GO and KEGG enrichment analysis. The differentially expressed genes from the disease dataset were then intersected with the ferroptosis-related genes to identify common genes. Core genes were selected using three machine learning algorithms, namely LASSO, SVM-RFE, and random forest. Further analysis included exploring immune infiltration correlation, predicting target drugs, and conducting molecular docking simulations.

Results: The differential analysis of the GSE45885 dataset yielded 1751 differentially expressed genes, while the GSE145467 dataset yielded 4358 differentially expressed genes. The intersection of these two gene sets resulted in a disease-related gene set consisting of 508 genes. Taking the intersection of the disease-related gene set and the ferroptosis-related gene set, we obtained 17 disease-related ferroptosis genes. After machine learning-based screening, three core genes were identified: GPX4, HSF1, and KLHDC3.

Conclusion: The mechanism underlying the involvement of ferroptosis in non-obstructive azoospermia may be linked to the downregulation of GPX4, HSF1, and KLHDC3 expression. This finding provides a basis for subsequent in-depth mechanistic and therapeutic studies.