Edinburgh Research Explorer

Probabilistic Inference Reveals Synapse-specific Synaptic Dynamics in Neocortical Microcircuits

Research output: Contribution to conferencePoster

Original languageEnglish
Publication statusPublished - 2013
EventComputational and Systems Neuroscience (Cosyne) 2013 - Salt Lake City, Utah, United States
Duration: 28 Feb 20135 Mar 2013

Conference

ConferenceComputational and Systems Neuroscience (Cosyne) 2013
CountryUnited States
CitySalt Lake City, Utah
Period28/02/135/03/13

Abstract

Short-term synaptic plasticity is highly diverse and varies with brain area, cortical layer, cell type, and develop-mental stage. Since this form of plasticity shapes neural dynamics, its diversity suggests a specific and essential
role in neural information processing. Therefore, a correct identification of short-term plasticity is an important step towards understanding and modeling neural systems. Accurate phenomenological models have been developed, but they are usually fitted to experimental data using least-mean square methods. We demonstrate that, for typical synaptic dynamics, such fitting gives unreliable results. Instead, we introduce a Bayesian approach based on a Markov Chain Monte Carlo method, which provides the posterior distribution over the parameters of the model. We test the approach on simulated data. First we show that common protocols to measure short-term plasticity protocols yield broad distributions over some model parameters, i.e. with inaccurate estimates. Using this insight, we find a better experimental protocol for inferring the true synaptic parameters and show that our Bayesian formulation provides robust identification of changes in the model parameters. Next, we infer the model
parameters using experimental data from three different neocortical excitatory connection types, revealing novel synapse-specific distributions, while the approach yields more robust clustering results. Our approach to demar-cate synapse-specific synaptic dynamics is an important improvement on the state of the art and reveals novel features from existing data as well as guiding future experimental work.

Event

Computational and Systems Neuroscience (Cosyne) 2013

28/02/135/03/13

Salt Lake City, Utah, United States

Event: Conference

ID: 17889092