MyoGen: Unified Biophysical Modeling of Human Neuromotor Activity and Resulting Signals

Abstract

Understanding human motor control requires integrating cortical and spinal cord activity, muscle mechanics, and electrophysiology, levels that are often studied separately. We present MyoGen, an open-source framework that unifies spinal circuitry, proprioceptive feedback, musculotendon dynamics, cortical activity, and multimodal electromyography (EMG) generation in a single, interoperable platform. Spinal motor neurons and the resulting motor unit (MU) action potentials represent the only neural cells that can be accessed at scale in humans. We used human MU ensembles to validate our model across a wide range of experimental conditions. Using data-driven optimization, we found that MyoGen produces MU population activity that closely matches experimental discharge-rate distributions and discharge variability across human muscles. Moreover, it generates decomposable surface and intramuscular EMG, reproduces beta-band modulation of descending drive and its nonlinear transformation into force, and implements complete sensorimotor loops. Dimensionality reduction of simulated agonist-antagonist EMG reveals low-dimensional control manifolds consistent with experimental recordings from both healthy and spinal cord injured individuals. MyoGen provides physiologically grounded, ground-truth data and integrates seamlessly with analysis pipelines, enabling systematic investigation of motor control principles, validation of signal-processing algorithms, and exploration of sensorimotor interactions that are experimentally inaccessible.