In Aging ; h5-index 49.0
Osteoarthritis (OA) is one of the most common diseases in the orthopedic clinic, characterized by progressive cartilage degradation. RNA-binding proteins (RBPs) are capable of binding to RNAs at transcription and translation levels, playing an important role in the pathogenesis of OA. This study aims to investigate the diagnosis values of RBP-related genes in OA. The RBPs were collected from previous studies, and the GSE114007 dataset (control = 18, OA = 20) was downloaded from the Gene Expression Omnibus (GEO) as the training cohort. Through various bioinformatical and machine learning methods, including genomic difference detection, protein-protein interaction network analyses, Lasso regression, univariate logistic regression, Boruta algorithm, and SVM-RFE, RNMT and RBM24 were identified and then included into the random forest (RF) diagnosis model. GSE117999 dataset (control = 10, OA = 10) and clinical samples collected from local hospital (control = 10, OA = 11) were used for external validation. The RF model was a promising tool to diagnose OA in the training dataset (area under curve [AUC] = 1.000, 95% confidence interval [CI] = 1.000-1.000), the GSE117999 cohort (AUC = 0.900, 95% CI = 0.769-1.000), and local samples (AUC = 0.759, 95% CI = 0.568-0.951). Besides, qPCR and Western Blotting experiments showed that RNMT (P < 0.05) and RBM24 (P < 0.01) were both down-regulated in CHON-001 cells with IL-1β treatment. In all, an RF model to diagnose OA based on RNMT and RBM24 in cartilage tissue was constructed, providing a promising clinical tool and possible cut-in points in molecular mechanism clarification.
Yin Wenhua, Lei Ying, Yang Xuan, Zou Jiawei
2023-Jan-05
RNA-binding protein, diagnosis, machine learning, osteoarthritis, random forest
