Path knockout: an analysis tool for connectome-driven circuit discovery

Abstract

Neural circuits in the brain transform sensory information into perceptual, emotional and motor output. The connectome -- a synaptic-resolution wiring diagram of all neurons in a nervous system -- provides a neuroanatomical blueprint for identifying different circuits that mediate specific computations. However, its sheer scale and complexity pose a significant challenge in extracting functionally relevant multi-synaptic pathways. Here we established path knockout, a connectome analysis method that identifies dominant multi-synaptic pathways. Path knockout uses the input fraction of each synaptic connection as a surrogate for the probability that information transfer succeeds across that synapse and calculates approximate probability, or connection strength, that at least one path succeeds among all multi-synaptic paths between defined source and sink neuron populations. It then quantifies the importance of each interneuron by assessing the impact of its removal on the connection strength. We demonstrate that path knockout outperforms other methods for pathway analysis and identifies known and novel interneurons in the Drosophila circuits. Using path knockout, we identified a convergent circuit motif linking sexually dimorphic aggression command neurons to a shared isomorphic visual pursuit pathway. Path knockout is implemented such that it runs in minutes on a standard computer. Thus, path knockout provides a scalable and generalizable framework for connectome-driven circuit discovery.