Osteoarthritis (OA) and rheumatoid arthritis (RA) are joint diseases that share similar clinical features but have different etiologies, making a differential diagnosis particularly challenging. Background/Objectives: Utilizing advanced machine learning (ML) techniques on metabolomic data, this study aimed to identify key metabolites in synovial fluid (SF) that could aid in distinguishing between OA and RA. Methods: Metabolite data from the MetaboLights database (MTBLS564), analyzed using nuclear magnetic resonance (NMR), were processed using normalization, a principal component analysis (PCA), and a partial least squares discriminant analysis (PLS-DA) to reveal prominent clustering. Results: Decision forests and random forest classifiers, optimized using genetic algorithms (GAs), highlighted a selection of a few metabolites—primarily glutamine, pyruvate, and proline—with significant discriminative power. A Shapley additive explanations (SHAP) analysis confirmed these metabolites to be pivotal predictors, offering a streamlined approach for clinical diagnostics. Conclusions: Our findings suggest that a minimal set of key metabolites can effectively be relied upon to distinguish between OA and RA, supported by an optimized ML model achieving high accuracy. This workflow could streamline diagnostic efficiency and enhance clinical decision-making in rheumatology.
Advanced Machine Learning for Comparative Synovial Fluid Analysis in Osteoarthritis and Rheumatoid Arthritis
Kopeć, Karolina Krystyna;Uccheddu, Gabrieleanselmo;Noto, Antonio;Piras, Cristina;Spada, Martina;Atzori, Luigi;Fanos, Vassilios
2025-01-01
Abstract
Osteoarthritis (OA) and rheumatoid arthritis (RA) are joint diseases that share similar clinical features but have different etiologies, making a differential diagnosis particularly challenging. Background/Objectives: Utilizing advanced machine learning (ML) techniques on metabolomic data, this study aimed to identify key metabolites in synovial fluid (SF) that could aid in distinguishing between OA and RA. Methods: Metabolite data from the MetaboLights database (MTBLS564), analyzed using nuclear magnetic resonance (NMR), were processed using normalization, a principal component analysis (PCA), and a partial least squares discriminant analysis (PLS-DA) to reveal prominent clustering. Results: Decision forests and random forest classifiers, optimized using genetic algorithms (GAs), highlighted a selection of a few metabolites—primarily glutamine, pyruvate, and proline—with significant discriminative power. A Shapley additive explanations (SHAP) analysis confirmed these metabolites to be pivotal predictors, offering a streamlined approach for clinical diagnostics. Conclusions: Our findings suggest that a minimal set of key metabolites can effectively be relied upon to distinguish between OA and RA, supported by an optimized ML model achieving high accuracy. This workflow could streamline diagnostic efficiency and enhance clinical decision-making in rheumatology.File | Dimensione | Formato | |
---|---|---|---|
metabolites-15-00112.pdf
accesso aperto
Descrizione: Articolo principale
Tipologia:
versione editoriale (VoR)
Dimensione
2.21 MB
Formato
Adobe PDF
|
2.21 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.