CRCNS US-Israel Research Proposal: Understanding single neuron computation by combining biophysical and statistical models

Carnegie Mellon University, Pittsburgh PA

Investigators

Abstract

What make a neuron become active? This question, central to our understanding of brain processes, is both a biophysical question about underlying biological mechanisms, and a statistical question about the features of incoming stimuli to which a neuron responds. The overall goal of this project is to forge a link between the biological mechanisms of neuronal activity and the computational process by which neurons encode features of incoming stimuli. More specifically, this project seeks to understand the biological underpinnings of stimulus coding by neurons. Dynamical models of neurons that incorporate detailed information about the ion channels that these cells express provide a detailed, biophysical account of neuronal activity. These kinds of models have been used widely and can incorporate and constrain an impressive amount of biological detail. Unfortunately they provide little insight into the meaning of neuronal activity or into the kinds of computations and transformations of stimuli that neurons are performing. On the other hand, models derived from statistical approaches are able to capture the often-noisy and complex relationships between neural activity and the stimuli that a neuron receives. These models provide insight into how specific features of incoming stimuli are extracted and combined by populations of neurons. These approaches will be combined through collaboration of a team at Carnegie Mellon University (Urban and Kass) and one at Bar-Ilan University (Korngreen) with expertise in the application of statistical and biophysical models to single neuron data. The work will focus on two neuron types that have several important features in common. Olfactory bulb mitral cells and layer 5 neocortical pyramidal cells are two classes of large neurons that receive distinct sources of input inputs onto different divisions of their elaborate dendritic trees. To forge this connection between dynamic and statistical models, this project will develop detailed biophysical models using recently described methods and extend the framework of current statistical models to allow the interpretation of the functional consequences of ion channels and their localization on specific classes of inputs. Applying these improved methods, and examining the consequences of changing biophysical properties on the ability of neurons to robustly and effectively represent stimuli will generate a novel account of the linkage because biological mechanisms and single neuron computation. A companion project is being funded by the Israel Binational Science Foundation (BSF).