Bioinformatics identification and validation of biomarkers and infiltrating immune cells in endometriosis

Abstract

BackgroundEndometriosis (EM) is a common gynecological disorder that often leads to irregular menstruation and infertility. The pathogenesis of EM remains unclear and delays in diagnosis are common. Thus, it is urgent to explore potential biomarkers and underlying molecular mechanisms for EM diagnosis and therapies.MethodsThree EM-related datasets (GSE11691, GSE25628, and GSE86534) were downloaded from the Gene Expression Omnibus (GEO) which were integrated into a combined dataset after removing batch effect. Differentially expressed immune cell-related genes were obtained by CIBERSORT, WGCNA, and the identification of differentially expressed genes. Random forest model (RF), support vector machine model (SVM), and generalized linear model (GLM) were then constructed and the biomarkers for EM were determined. A nomogram evaluating the risk of disease was constructed and the validity was assessed by the calibration curve, DCA curve, and clinical impact curve. Single-gene Gene Set Enrichment Analysis (GSEA)was performed to explore the molecular mechanisms of biomarkers. The ceRNA regulatory network of biomarkers was created by Cytoscape and potential target drugs were obtained in the DGIdb database (Drug-Gene Interaction database).The expression levels of biomarkers from clinical samples was quantified by RT-qPCR.ResultsThe ratio of eight immune cells was significantly different between the eutopic and ectopic endometrium samples. A total of eight differentially expressed immune cell-related genes were investigated. The SVM model was a relatively suitable model for the prediction of EM and five genes (CXCL12, PDGFRL, AGTR1, PTGER3, and S1PR1) were selected from the model as biomarkers. The calibration curve, DCA curve, and clinical impact curve indicated that the nomogram based on the five biomarkers had a robust ability to predict disease. Single gene GSEA result suggested that all five biomarkers were involved in labyrinthine layer morphogenesis and transmembrane transport-related biological processes in EM. A ceRNA regulatory network containing 184 nodes and 251 edges was constructed. Seven drugs targeting CXCL12, 49 drugs targeting AGTR1, 16 drugs targeting PTGER3, and 21 drugs targeting S1PR1 were extracted as potential drugs for EM therapy. Finally, the expression of PDGFRL and S1PR1 in clinical samples was validated by RT-qPCR, which was consistent with the result of public database.ConclusionsIn summary, we identified five biomarkers (CXCL12, PDGFRL, AGTR1, PTGER3, and S1PR1) and constructed diagnostic model, furthermore predicted the potential therapeutic drugs for EM. Collectively, these findings provide new insights into EM diagnosis and treatment.

LGALS2 and EGR1: markers of endometriosis for predictive, preventive and personalized medicine

“…CLINT1 is involved in hormone activity, fat-soluble vitamin metabolic process, and protein-DNA complex. CLINT1 is involved in hormone activity, fat-soluble vitamin metabolic process, protein-DNA complex subunit organization, protein serine kinase activity, and neuroactive ligand-receptor interaction.This is similar to previous research [47][48][49] and also corresponds to the previous research of our group [50] .Overall, the three core molecules screened may be involved in the efferocytosis process of EM from different perspectives, which in turn affects the development of EM.…”

Bioinformatics-Based Analysis of Key Genes in Steroid-Induced Osteonecrosis of the Femoral Head That Are Associated with Copper Metabolism

“…To study the molecular mechanisms associated with the diagnostic key genes, GSEA software (V4.0.3) [ 29 ] was utilized to perform single-gene GSEA [ 30 ]. When the parameter “c2.cp.kegg.v7.4.symbols.gmt” was set, the KEGG pathway gene set was used as the enrichment background [ 22 , 23 , 24 ]; when the parameter “c5.go.bp.v7.4.symbols.Gmt” was set, the GO biological process gene set was set as the enrichment background; when the parameter “Phenopyte labels: Use a gene as the phenotype” was used, the expression values of two key genes were considered the phenotype file.…”

Combining bioinformatics and machine learning to identify common mechanisms and biomarkers of chronic obstructive pulmonary disease and atrial fibrillation

“…Use 70% of the data for model construction and 30% for model validation. Models for RF, SVM, XGBoost, and GLM were constructed separately using the “caret” R package ( 25 ). It is well known that while building a machine learning model with good results, it is equally important to evaluate the interpretation of the model, as only an interpretable machine learning model is likely to be more widely understood and adopted.…”