Related Experiment Video
Updated: Aug 8, 2026
Performing Data Mining And Integrative Analysis Of Biomarker in Breast Cancer Using Multiple Publicly Accessible Databases
Published on: May 17, 2019
Integrated transcriptomic and machine learning analysis identifies female-associated candidate biomarkers in
Atefeh Pourdadashi1, Pariya Eskandari2, Zahra Ghayour Najafabadi3
1Department of Biology and Anatomical Science, School of Medicine, Shahid Beheshti University of Medical Sciences, Tehran, Iran.
Background:
Idiopathic pulmonary arterial hypertension (IPAH) exhibits a marked female predominance. Despite this "estrogen paradox," the molecular basis of sex-associated pathogenesis remains incompletely understood, limiting the development of precision diagnostic strategies.
Objective:
To develop a sex-stratified computational framework that integrates transcriptomic profiling and ensemble machine learning to identify female-associated candidate biomarkers in IPAH.
Methods:
Sex-stratified differentially expressed genes (DEGs) were derived from pulmonary arterial endothelial cells (GSE303084, n = 48) and lung tissue (GSE126262, n = 16). Functional mechanisms were explored via GO/KEGG enrichment and protein-protein interaction (PPI) networks. A consensus machine learning framework incorporating 12 feature selection algorithms prioritized stable candidate biomarkers. Model performance was evaluated using a leakage-free repeated stratified 5-fold cross-validation framework, followed by independent external validation in a female IPAH cohort. Sex-associated immune landscapes were characterized via CIBERSORTx-based deconvolution.
Results:
Female-associated DEGs were associated with metabolic dysregulation, lysosomal pathways, and TGF-β/Wnt signaling. PPI analysis identified MMP1, SFRP1, SNAI2, ADAM12, and INHBA as central hub genes. The ensemble framework prioritized METTL27 and CDH11 as candidate biomarkers. A two-gene signature (METTL27 + CDH11) demonstrated robust performance under leakage-free repeated cross-validation (XGBoost mean AUC 0.875 ± 0.161) and correctly identified 26 of 28 female IPAH patients in an independent external cohort (patient-detection sensitivity 92.9%; 95% CI, 77.4-98.0%). Immune deconvolution revealed marked sex dimorphism, with adaptive immune enrichment in females and innate myeloid predominance in males.
Conclusion:
Our approach identifies a candidate two-gene signature (METTL27/CDH11) for female IPAH and highlights TGF-β/Wnt signaling, metabolic, and immune dimorphism as key drivers of sex-associated pathogenesis. These findings provide a foundation for future clinical and experimental validation of sex-informed candidate molecular biomarkers in IPAH.