Modeling the Neural Activity of Caenorhabditis Elegans Through Neural Message Passing

Abstract

The great complexity of the human connectome motivates the study of a simpler neural network. For that purpose, the Ising Model was applied on experimental data on the synaptic connectivity of Caenorhabditis elegans (C. elegans) in resting-state, assigning a binary variable (representing active or inactive states) to each neuron in the network. The dynamics of this system is postulated as a message passing network, encoded by the Belief Propagation algorithm (BP) in its criticality region. The inferences of neuronal activity maps were obtained for different times of the nematode’s life cycle. We determined the network susceptibilities as a measure of correlations in the system through the Susceptibility Propagation algorithm (SP). Finally, we applied clustering methods to obtain functional clusters and analyze similarities between them and the real functional clusters (sensory, interneurons and motor). All this contributed to the analysis of structure-function relationship in the C. elegans neural network.