Work Package: WP5
Deliverable: 11.2: Neural decoding from 200-neuron patterns with development of a Matlab toolbox
Deliverable due date: month 24
This document reports on the progress of work on Deliverable 11.2. The planned tasks have been completed - they have yet to be applied on experimental data. We recently submitted our results to “PloS Computational Biology” under the Open Access policy, which will makes it accessible with no restrictions. A pre-print version of this article is already available on BioRxiv (http://biorxiv.org/content/early/2016/01/26/023200). This work has been illustrated by two animations demonstrating how activity in large population of presynaptic neurons (<600 neurons) impact on a post-synaptic neuron depending on the spatial distribution of their synapses. The code used to produce these animations will be available upon acceptance of our manuscript. It will be uploaded on the git repository of the main author (https://github.com/rcaze). This code can be used to better decode patterns produced by large ensembles of neurons (<200)
In the previous study, we simulated activity in 8 ensembles 250 neurons. Each ensemble is composed of a correlated population of neurons and we show how these correlations can produce rich activity patterns and give rise to stimulus selectivity. We made a neuron stimulus selective by playing on the spatial distribution of active synapses. The preferred stimulus makes dispersed contacts while non-preferred stimuli make clustered synapses on the post-synaptic neuron. The methodological tool we used in this study can be used to understand neural recordings from large population of neurons. Therefore this work fosters a new understanding of large neural recordings and can be used to understand the observed pattern of activity.
In such a way we fulfilled Deliverable 11.2: “Neural decoding from 200-neuron patterns with development of a Matlab toolbox”. As with deliverable 11.1, we used Python instead of Matlab to implement our algorithms. While Matlab need to be purchased Python is freely accessible to anyone, this will make our work usable in other studies.
The work also covers milestone 12.
Contributors: Romain Caze, Sarah Jarvis, Amanda Foust, Simon Schultz (Imperial)