Abstract
Facioscapulohumeral muscular dystrophy (FSHD) is an autosomal dominant muscle disorder characterized by a complex genetic etiology, variable prognosis, and a lack of effective therapies. Previous studies have identified candidate protein and miRNA biomarkers using various profiling techniques, underscoring their potential for monitoring FSHD, assessing prognosis, and evaluating pharmacodynamic responses. However, the feasibility of applying machine learning (ML) models to predict FSHD using these molecular signatures has not been explored. In this study, we developed ML models to predict FSHD using a multi-omics dataset comprising protein abundance and miRNA expression profiles. Key predictive features were identified using Random Forest and the Support Vector Machine-Recursive Feature Elimination (SVM-RFE) methods. Performance evaluations demonstrated the robustness of the ML classifiers, with logistic regression consistently achieving the robust predictive accuracy in distinguishing FSHD from healthy conditions. Additionally, we assessed the predictive power of the identified features by comparing them with biomarker sets reported in previous studies. Our findings highlight the potential of AI to improve prediction accuracy and facilitate the cost- and time-efficient strategy for identifying FSHD biomarker candidates, even with limited sample sizes in the context of rare diseases.