A Compressed Sensing Framework for Efficient Dissection of Neural Circuits

A fundamental question in neuroscience is how neural networks generate behavior. The lack of neuron subtype specific genetic tools makes it challenging to determine the roles of individual subtypes in behavior. We describe a compressed sensing based framework to rapidly infer candidate neurons controlling behaviors with much fewer measurements than previously thought possible by exploiting non-specific genetic tools. We tested this framework by inferring interneuron subtypes regulating the speed of locomotion of the nematode Caenorhabditis elegans. We validated the inferences using a novel real time stabilization microscope for accurate long-time, high magnification imaging and targeted perturbation of neural activity in freely moving animals. We show that a circuit of three interconnected interneuron subtypes, RMG, AVB and SIA control different aspects of the speed as the animal navigates the environment. Our work suggests that compressed sensing approaches can be broadly used to identify key nodes in complex biological networks.