Machine Learning Classification of Serum miR-210 as a Biomarker of Exercise Training Response

Area of Science:

• Exercise physiology
• Biomarker discovery
• Machine learning in healthcare

Background:

• Circulating microRNAs (miRNAs) are emerging biomarkers for exercise adaptation.
• miR-210, linked to hypoxia, angiogenesis, and stress, shows variable exercise responses.
• The potential of miR-210 as a classifier for individual training responses remains unestablished.

Purpose of the Study:

• To evaluate the utility of serum miR-210 as a biomarker for individual exercise training responses.
• To assess the performance of machine learning models in classifying miR-210 upregulation post-exercise.

Main Methods:

• Ten young adults underwent standardized exercise training.
• Serum miR-210 expression was quantified via qRT-PCR.
• Three machine learning models (ridge-penalized logistic regression, random forest, SVM) were employed with Leave-One-Out Cross-Validation.

Main Results:

• Machine learning classifiers achieved 90% accuracy in identifying miR-210 upregulation.
• Random Forest classifier demonstrated superior performance with high AUC, sensitivity, and specificity.
• No significant correlations were found between miR-210 levels and training frequency or age.

Conclusions:

• Machine learning models accurately identify serum miR-210 upregulation post-exercise.
• Serum miR-210 shows potential as a discriminative biomarker for inter-individual exercise responses.
• These preliminary findings require validation in larger cohorts to establish clinical predictive value.