Bridging the gap between the connectome and whole-brain activity in C. elegans

A fundamental goal of neuroscience is to understand how anatomy determines the functional properties of the nervous system. However, it has been challenging to relate large-scale functional measurements of the C. elegans nervous system to the worm’s known anatomical connectome ^1–3 . Here, we address this apparent discrepancy using a connectome-constrained model of the nematode brain fit to neural recordings with optogenetic perturbations ² . Our model consists of a noisy linear dynamical system with a sparse synaptic weight matrix with non-zero entries only where there are synapses in the C. elegans connectome. We evaluated the model by perturbing neurons in silico and measuring the perturbation response of all other neurons in the network. We compared these responses to those measured in held-out animals and found that this model captured the perturbation-triggered responses of individual neurons 92% as well as the reproducibility of the perturbation responses themselves. This includes perturbation responses of neurons that were not anatomically connected, which the model explains in terms of signal propagation over multiple neurons. In addition to capturing perturbation responses, the model also accurately predicts the activity of held-out neurons using the observed activity of other neurons. Strikingly, alternative models with equivalent levels of sparsity but a shuffled connectome constraint achieved much lower performance. Finally, we demonstrate that adding connections beyond those found in the connectome did not improve the model’s prediction of the perturbation measurements. The model described here provides the strongest link yet between the connectivity of the C. elegans nervous system and its causal and correlative functional properties.