Statistical Inference of the Cerebellar Network

ended

The brain can coordinate complex sequences of actions with the accuracy of milliseconds. Where and how these neural computations occur is an open question in neuroscience. Despite recent technological developments allowing for large-scale high-resolution functional imaging of the brain and direct neuronal recordings in behaving animals, there has been little effort in applying rigorous statistical approaches to test circuit connectivity patterns and synaptic mechanisms driving neural activity. Experimental evidence from classical conditioning and neuronal recordings have revealed that the cerebellum plays a fundamental role in fine-tuning of temporally precise behaviors. This project aims to elucidate the neural computation arising from anatomical and physiological constraints of the comparatively simple organization of the cerebellar cortical circuit, which allows the cerebellum to represent time-dependent sensory information necessary to drive behavior. Experimental and theoretical findings in the host laboratory have led to the hypothesis that dynamic synapse are a substrate for temporal representations and temporal learning. I will use sequential Monte Carlo methods to extract activity from calcium imaging data. Then I will use a generative model of the cerebellar network to infer the connectivity among the known cell types of the cerebellum as well as their synaptic properties. Finally, I will use information theory to examine the processing capacity of the cerebellar network, thereby providing new insights on evolutionary optimization of brain computation. The combination of my experience in statistical methods and the host laboratory's experience in state-of-art neural recordings and theoretical models, is a perfect match to break down the barriers to understanding the cellular mechanisms of circuit computations. We believe that this analysis approach could also be applied to understand other neuronal circuits.