Machine Learning Identified Mitophagy- and Hypoxia-Related Genes MTRNR2L8 and SUMO1P3 as Potential Biomarkers Associated with Unexplained Recurrent Pregnancy Loss.

This study aims to identify mitophagy- and hypoxia-related biomarkers for unexplained recurrent pregnancy loss (uRPL) by integrating transcriptomic data with machine learning algorithms. The endometrial transcriptome dataset GSE165004 was obtained from the GEO database to identify differentially expressed genes (DEGs) between uRPL cases and healthy controls. A set of mitophagy- and hypoxia-related genes (HMRGs) was collected from GeneCards and PubMed. Weighted gene co-expression network analysis (WGCNA) was used to detect HMRG-related key module genes. The intersection between module genes and DEGs was defined as DE-HMRGs, which were subjected to functional and pathway enrichment analysis and protein-protein interaction network (PPI) construction. Hub genes were selected using LASSO, SVM-RFE, and random forest algorithms, and their diagnostic performance was evaluated by an artificial neural network (ANN) model and validated in an independent dataset. A total of 356 DEGs were identified. Intersecting these DEGs with HMRG key module genes yielded 82 DE-HMRGs, which were mainly enriched in energy metabolism and WNT signaling. In the PPI network constructed from DE-HMRGs, NCKAP1 emerged as a central node. Machine learning identified MTRNR2L8 and SUMO1P3 as key hub genes, exhibiting good diagnostic performance in the ANN model in both the training and validation sets (AUC=1, respectively). The mitophagy- and hypoxia-related genes MTRNR2L8 and SUMO1P3 are involved in the pathogenesis of uRPL and may serve as potential diagnostic and therapeutic biomarkers. However, the study is limited by its retrospective nature and a lack of experimental validation; thus, prospective studies and functional validation are warranted.