OP2-P700 DRIVING KNEE OSTEOARTHRITIS STRUCTURAL PROGRESSOR PROGNOSIS INTO THE NEXT GENERATION: LEVERAGING MACHINE/DEEP LEARNING, MICRORNA AND MAGNETIC RESONANCE IMAGING

Document Type

Conference Proceeding

Publication Date

9-23-2025

Publication Title

Aging Clin Exp Res

Keywords

biological marker, microRNA, microRNA 200a, adult, African American, artificial neural network, Caucasian, cohort analysis, conference abstract, controlled study, deep learning, diagnostic test accuracy study, dimensionality reduction, drug therapy, elastic tissue, feature selection, female, human, knee, knee osteoarthritis, learning algorithm, machine learning, major clinical study, male, middle aged, MRI scanner, nuclear magnetic resonance imaging, osteoarthritis, predictive model, prognosis, X ray analysis

Abstract

Objective: Predicting knee osteoarthritis (OA) patients at risk of rapid structural progression remains challenging. Circulating micro-RNAs (miRNAs) showed promise as biomarkers for stratifying such patients. This study aimed to develop a miRNA-based prognosis model to identify knee OA structural progressors using machine/deep learning, with structural progressors defined via a methodology using MRI and X-ray data1. Methods: Baseline serum miRNAs from 152 Osteoarthritis Initiative (OAI) participants were isolated, sequenced, and used for model development. Dimensionality reduction was performed to identify the most informative miRNAs within the initial set of 456. Key miRNAs and OA determinants, including age, sex, BMI, and race (Caucasian and African American), were selected after a comprehensive exploration of 7 feature selection machine learning. The final model was developed after extensively exploring an array of machine/deep learning algorithms. The performance of the models was assessed using AUC, accuracy, sensibility, and specificity. Validation employed an independent cohort of 30 OAI baseline plasma samples. Results: Feature clustering reduced the initial set to 107 miRNAs. Elastic Net was identified as the optimal feature selection model. The final prediction model utilized an Artificial Neural Network (ANN) comprised of age and four miRNAs, hsa-miR-556-3p, hsa-miR-31575p, hsa-miR-200a-5p, and hsa-miR-141-3p, and achieved an excellent performance (AUC, 0.94; accuracy, 0.84; sensitivity, 0.89; specificity, 0.75). The ANN model validation analysis confirmed the model's robustness (AUC, 0.81; accuracy, 0.83; sensitivity, 0.71; specificity, 0.94). Conclusion: This study identifies, for the first time, a microRNA signature capable of predicting rapid structural progression in knee OA patients. The model demonstrated strong performance and was validated in an independent cohort, showcasing its potential for generalization. The translational potential of this prediction model is significant, as it will provide clinicians with a valuable tool for personalized and targeted treatment strategies as well as assist early identification of high-risk structural progressor patients for inclusion in trials.

Volume

37

Issue

1

First Page

279

Find It @ Sladen

Share

COinS