Causal pattern recovery (CPR) - a novel analysis technique for neural spike train data
-
1
University of St Andrews, United Kingdom
-
2
Newcastle University Medical School, United Kingdom
Electrophysiological brain recordings are a valuable data source for both experimentalists and modellers. In order to utilise the large amounts of multi-channel spike train data from such recordings, powerful computational tools are needed to extract the critical features for further analysis. CPR is a novel computational analysis technique, which detects stochastic dependencies between data channels to support building new models and refinement of existing ones.
CPR learns a probabilistic graphical model (Dynamic Bayesian Network) from the multi channel spike train data using a novel scoring function, which combines a non-linear filter and spike triggered snapshots of its output. The scoring function is used with machine learning techniques to check numerous potential causal networks, where networks with high causal quality will be assigned a high score value. Stochastic dependencies can then be derived from the identified top-scoring networks to serve as a valuable basis for sophisticated models.
Preliminary tests with real data confirm the applicability of CPR: We applied CPR to data collected by Evelyne Sernagor and Chris Adams, who used a high density multi-electrode array to record spontaneous activity waves from the neonatal mouse retina. We successfully revealed dependency-networks that depict the path of the travelling wave well (Fig. 1). Our ongoing research includes assessing CPR’s strengths and weaknesses under controlled conditions using a comprehensive series of simulations. In each step of the series we generate a random network that defines synaptic connections between individual neurons. We then simulate the network using different kinds of neuron models, to generate synthetic spike train data to which CRP is applied. In the final step, the stochastic relations revealed by CPR are employed to fit a model of the observed system to the data; The model’s prediction quality is then compared against its quality when fitted using the true stochastic relations. With this framework we can assess the benefit of CPR for complex models, get valuable insights into the method’s capability and estimate its range of potential application.
Initial considerations suggest that the applicability of CPR is not limited to electrophysiological recordings alone but that it may also be applied to other kinds of physiological data such as calcium imaging, for example.
The development of CPR is part of the CARMEN-project (Code Analysis, Repository and Modelling for e-Neuroscience) and CPR will become publicly available as a tool on the CARMEN-platform (www.carmen.org.uk).
Conference:
Neuroinformatics 2008, Stockholm, Sweden, 7 Sep - 9 Sep, 2008.
Presentation Type:
Poster Presentation
Topic:
Computational Neuroscience
Citation:
Echtermeyer
C,
Sernagor
E,
Adams
C and
Smith
V
(2008). Causal pattern recovery (CPR) - a novel analysis technique for neural spike train data.
Front. Neuroinform.
Conference Abstract:
Neuroinformatics 2008.
doi: 10.3389/conf.neuro.11.2008.01.016
Copyright:
The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers.
They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.
The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.
Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.
For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.
Received:
25 Jul 2008;
Published Online:
25 Jul 2008.
*
Correspondence:
Christoph Echtermeyer, University of St Andrews, St Andrews, Fife, United Kingdom, ce86@st-andrews.ac.uk