Neuronal Systems and Modeling

1244-Pos

Computational Neuroinformatic Toolkit I: Information-Theoretic Analysis of Spike Trains

Halls A/B/C/D

5M Neuronal Systems & Modeling

David H. Goldberg, Ph.D.¹, Jonathan D. Victor, M.D., Ph.D.², Daniel Gardner, Ph.D.¹.
¹Lab of Neuroinformatics, Weill Cornell Medical Coll, New York, NY, USA, ²Dept. of Neurology, Weill Cornell Medical Coll, New York, NY, USA.

Computational neuroinformatics synthesizes computational neuroscience--analyses of neural representation and information processing--and neuroinformatics--standards-based methods for archiving, classifying, and exchanging neuroscience data. We are assembling a computational neuroinformatic resource that will aid the investigation of neural coding. This resource consists of neurodatabase.org, a web-based neurophysiology data archive, and a collection of information-theoretic spike train analysis tools. By providing experimenters with advanced spike train analysis techniques and theoreticians with broad collections of datasets, we seek to advance collaborations toward understanding neural codes.
We now report a key component of this resource: an open source, portable toolkit for information-theoretic analysis of spike trains. The toolkit is implemented in C and includes a Matlab interface via the MEX framework. It contains implementations of three established methods: direct, metric space, and binless. Also included is a shared module that estimates entropy from a discrete histogram, a computation common to many information-theoretic methods. This module provides both classical and recent methods for estimating the bias and variance of entropy estimates. In an initial test, the toolkit was applied to the analysis of single neuron responses in macaque posterior parietal cortex during a prehension task.
The toolkit is being ported to a dedicated parallel cluster, enabling computationally intensive analyses such as those involving populations of simultaneously recorded neurons. The cluster will be accessible through and applicable to data archived at neurodatabase.org. These resources, along with midlayer components and a graphical interface, will comprise a web-enabled research tool for neuroscience data discovery.
Supported by Human Brain Project/Neuroinformatics MH068012 from NIMH, NINDS, NIA, NIBIB, and NSF.

 D.H. Goldberg, None; J.D. Victor, None; D. Gardner, None.