Statistical modelling of neural population activity in the mouse cortex across development

Lead Research Organisation: University of Bristol
Department Name: Computer Science

Abstract

The brain represents and transmits information as electrical pulses across populations of its
neurons. One of the central goals of neuroscience is to understand exactly how this
information is represented and transmitted. These days, researchers routinely record from
up to thousands of neurons simultaneously, even in animals that are awake and active.
However, despite this explosion in recording technologies, analysis methods capable of
properly quantifying information representation and transmission in large populations have
not developed at the same rate. The methods that have been developed each have their
own strengths and weaknesses. Given that the
size of the datasets created by experimental recordings contiues to grow, the need for new
analysis
techniques that can scale with the size of the neural population is clearer than ever at the
present time.
There are three main objectives to this project. The first is to develop a new analysis method
for estimating mutual information between a stimulus and the response of large populations
of neurons to that stimulus, which is scalable with the size of the neural population. The
second aim of this project is to compare a number of statistical models incuding the one
developed in this project when applied to data taken from the mouse somatosensory "barrel
cortex". Applying the models to the same data will allow a quantitive comparison of the
models. The third aim of this project is to build a neural network model of the mouse barrel
cortex based on data from the literature and constrained by the in vivo recorded data used
for the first parts of the project.
Initial development of a new analysis method will consist of extending the recently
developed population tracking model[1]. This method involves performing Bayesian
inference given probability distributions for population activation, and individual cell
activation. But methods of bias corrections will need to be developed. The limited sampling
bias, and the bias introduced by using a maximum entropy model will both have to be
corrected. Ideally, this method will be applicable to any neural population, but it may be
necessary to restrict the project to a certain neural population during the research period.
This new analysis method and other commonly used methods will be applied to Ca2+
imaging data taken from the mouse barrel cortex, and the monkey visual cortex. The results
of the analysis methods will be compared quantatively using parameters such as mutual
information. A neural network model of the mouse barrel cortex will be built. Furthermore, it
is aimed to use this model to address the question of how cortical population coding of
whisker stimuli changes during brain development. The analysis methods used in the earlier
objectives should play a role in achieving this aim.

Publications

10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/N509619/1 01/10/2016 30/09/2021
1793055 Studentship EP/N509619/1 19/09/2016 30/06/2020 Thomas Delaney
EP/R51245X/1 01/10/2017 30/09/2021
1793055 Studentship EP/R51245X/1 19/09/2016 30/06/2020 Thomas Delaney