Recurrent Implantation Failure: Bioinformatic Discovery of Biomarkers and Identification of Metabolic Subtypes

Abstract

Recurrent implantation failure (RIF) is a challenging scenario from different standpoints. This study aimed to investigate its correlation with the endometrial metabolic characteristics. Transcriptomics data of 70 RIF and 99 normal endometrium tissues were retrieved from the Gene Expression Omnibus database. Common differentially expressed metabolism-related genes were extracted and various enrichment analyses were applied. Then, RIF was classified using a consensus clustering approach. Three machine learning methods were employed for screening key genes, and they were validated through the RT-qPCR experiment in the endometrium of 10 RIF and 10 healthy individuals. Receiver operator characteristic (ROC) curves were generated and validated by 20 RIF and 20 healthy individuals from Peking University People’s Hospital. We uncovered 109 RIF-related metabolic genes and proposed a novel two-subtype RIF classification according to their metabolic features. Eight characteristic genes (SRD5A1, POLR3E, PPA2, PAPSS1, PRUNE, CA12, PDE6D, and RBKS) were identified, and the area under curve (AUC) was 0.902 and the external validated AUC was 0.867. Higher immune cell infiltration levels were found in RIF patients and a metabolism-related regulatory network was constructed. Our work has explored the metabolic and immune characteristics of RIF, which paves a new road to future investigation of the related pathogenic mechanisms.