Neural computation in the context of upstream dynamics in the retina

Abstract

To understand neural circuit function, one would like to understand individual neurons' computation in the context of their physiological inputs. But these inputs are themselves subject to complex dynamics, and we rarely have tools to experimentally control synaptic inputs under physiological conditions that preserve their temporal features. Primary sensory structures present an exception because primary receptor neurons can be controlled experimentally under physiological conditions. Here, I present work carried out in the retina, where signaling by photoreceptors (the primary receptor neurons in vision) has been characterized in detail and can be controlled with light, and computation in downstream circuitry can therefore be investigated in the context of physiological input from photoreceptors. I first present a hybrid biophysical-statistical model of retinal output that disentangles the computational contributions of photoreceptors from those of other circuit elements and successfully predicts retinal ganglion cell responses to stimuli with dynamically changing statistics. Second, I present an investigation of synaptic specializations that could mediate parallel processing of input from different photoreceptor types within individual post-synaptic neurons.