Thinking Dots

Visualizing dynamic neural networks

Computing the activity in the entire brain of a worm

 
 

We mapped causal signal flows in the complete nervous system (the connectome) of the worm C. elegans in response to the activation of a feeding circuit. Starting from a pattern of events (spikes) at sensory neurons that respond to the presence of food, we inferred all the signaling patterns that lead to alternating activations of populations of motor neurons responsible for locomotion. By analyzing the temporal sequences of signaling patterns we identified fundamental topological features in the dynamics of the network responsible for persistent equilibrium steady state activity - stable encodings of input data into the network. While we had to make some simplifying assumptions in our models due to gaps in our knowledge about the neurobiology,  these visualization give an intuitive feel for the dizzying types of dynamics and activity going on within the connectome.

 
 

We modeled the dynamics on the network based on a competitive refractory model derived from signaling in the biological brain, whereby arriving signals from multiple input neurons (nodes) into a common target node compete to activate the target node. But the target neuron is capable of exhibiting a refractory state. When the target neuron is not refractory the first arriving signal successfully activates it and causes it to fire. But when it is refractory, an arriving signal is not able to make the neuron fire. Therefore, the timing of when signals arrive from upstream neurons becomes critical to the dynamics - the behavior- of the network. (Green nodes represent excitatory neurons and red neurons inhibitory neurons.)
 

 
 

Movie of dynamic network flows through a C elegans feeding circuit.

Movie of network dynamic activity throughout the entire C elegans connectome in response to activation of the feeding circuit. 


Screenshot 2017-09-17 16.41.11.png