Prediction and validation of in vivo dendritic processing

ended

Integration of synaptic input by single neurons is fundamental to computation in the brain. The output of every cell within a network is shaped by the elaborate morphology of its dendritic tree, and a suite of biophysical mechanisms that confer nonlinear processing capabilities. Over past decades, a remarkable synergy between theory and experiment has elucidated key strategies by which single-cell processing could thus enhance the neural code. However, a critical gap in current understanding remains: which operations from the vast dendritic repertoire are actually employed in vivo? One major obstacle to addressing this problem is that the statistics of in vivo input patterns are largely unknown. Thus, it is unclear how salient information is presented to the dendrites of a neuron, and by extension, what mechanisms are used for its transduction to action potential output at the axon. I aim to answer these questions by combining my expertise with that of the host lab to formulate theoretical predictions and then validate them with in vivo experiments. Specifically, I will construct a computational model of a cortical neuron that learns to discriminate synaptic input patterns, and use it to discover the optimal scheme for encoding and decoding information. I will thus predict the spatiotemporal patterns of synaptic input to stimulus-tuned neurons, and the biophysical mechanisms through which this information can be extracted. I will then test these predictions in primary visual cortex of awake behaving mice through two-photon dual-colour imaging of presynaptic glutamate release and postsynaptic calcium dynamics. By relating the algorithmic and biological function of neurons in the living brain, I anticipate this project will yield important insights into general principles of neural computation.