Decoding locomotion from population neural activity in moving <i>C. elegans</i>

We investigate neural representations of locomotion by recording calcium activity from the majority of neurons in the compact brain of the nematode C. elegans as it crawls freely. We find neurons tuned to features of the animal’s spontaneous locomotion, such as its velocity and gross body curvature. We developed a population decoder using linear ridge regression that predicts properties of the animal’s current locomotion from its population neural activity. We find that our population decoder outperforms best single neuron models for both velocity and curvature and captures a wider range of behavioral outputs. We also labeled the AVA neuron pair and investigated its role in the population. Finally we studied differences between population neural activity in the same worm when it is moving compared to immobilized. We find an increase in the magnitude of the correlation of the neurons during immobilization and recover previously reported stereotyped neural trajectories.