Machine-learning classification of motor unit types in the adult mouse

Area of Science:

• Neuroscience
• Muscle Physiology
• Computational Biology

Background:

• Motor unit diversity is crucial for muscle function but poorly defined in mice.
• Current mouse motor unit classification relies on subjective thresholds, lacking quantitative electrophysiological links.

Purpose of the Study:

• To quantitatively define mouse motor unit types by correlating motoneuron electrophysiology with muscle contractile properties.
• To develop a predictive model for motor unit classification using machine learning.

Main Methods:

• In vivo intracellular recordings of motoneuron electrophysiology.
• Unsupervised clustering and supervised machine learning (multinomial logistic regression) on 40 contractile variables.
• Feature selection to identify key electrophysiological predictors.

Main Results:

• Four distinct motor unit types (S, FR, FI, FF) were objectively identified via clustering.
• Motoneuron electrophysiological properties systematically varied across these identified motor unit types.
• A predictive model accurately classified motor unit types, with four key electrical features identified.

Conclusions:

• Establishes quantitative criteria linking mouse motoneuron excitability to muscle contractile phenotype.
• Provides a framework for incorporating motor unit diversity into neuromuscular research.
• Facilitates future investigations into neuromuscular physiology and disease in mice.