Neuroinformatics: Introduction
J. van Pelt, Netherlands Institute for Brain Research
Neuroinformatics is a new field of research at the intersection of the neurosciences and the computer / information sciences. Realizing the challenges faced by the international scientific community of managing, analyzing and sharing the vast quantities of highly diverse data produced by brain research the OECD Global Science Forum installed in 1996 a Working Group Neuroinformatics to formulate recommendations for a global effort in Neuroinformatics.
At 29-30 January 2004, the OECD Ministerial Committee for Scientific and Technological Policy endorsed the recommendations of this Working Group Neuroinformatics, that the management and exploitation of data about the brain can be best achieved through a co-ordinated, multidisciplinary, international effort. It has recommended the establishment of a new global mechanism, the "International Neuroinformatics Coordinating Facility" (INCF), creation of an associated funding scheme, the "Programme in International Neuroinformatics", and the establishment of national nodes and research programmes in neuroinformatics.
In this Introduction these recommendations will be discussed as well as the implications and opportunities for a Neuroinformatics Program in The Netherlands
Simulation studies on the generation of epileptic seizures and spreading
depression in neurons.
G.G.Somjen, Duke University, Durham USA and SILS, University of Amsterdam
The firing pattern of neurons is determined by the total repertoire of
their ionic membrane currents. Under normal circumstances the level of
activity is such that it rarely leads to substantial changes in the ionic
composition of the intracellular or the extracellular volume compartments.
However, conditions that result in high frequency firing or strong
synchronous activity can change at least temporarily the ion levels in
these compartments. We modeled the dynamics of membrane potential, the
underlying ion currents, and ionic composition of a neuron -- extracellular
space -- glia system in order to understand the conditions under which
epileptic seizure can arise or spreading depression evoked. The model was
implemented in the public domain NEURON simulation environment (Hines). The
known sodium, potassium and chloride currents in combination with active
Na/K and Ca extrusion mechanisms in the neuron and the glia compartment are
sufficient to produce the various types of epileptic seizures. The
conditions that lead to the all-or-none spreading depression event could
also be characterized.
NeuroInformatics of neuronal information processing: from Neural Networks to Neuro-imaging
C.Gielen
Dept. of Biophysics, University Nijmegen
The complexity of neurons and that of neuronal information processing by discrete events (action potentials) rather than by continuous signals has been a source of debate since long: is the complexity essential for providing the unique and unsurpassed abilities of the brain, or is it an artefact of the biological substrate that is a handicap rather than an advantage. An answer to this question has come within reach thanks to new tools in theoretical and computational neuroscience and thanks to new experimental techniques.
The standard model for a neuron has been that of the leaky integrate-and-fire (IF) neuron. It has been a favorit model because it allows an analytical approach to investigate the dynamics of an ensemble of interacting neurons. One of the properties of the leaky IF neuron is that the coefficient of variation (CV, a measure of the variability of firing) increases with firing rate. Recently, it was shown that the Hodgkin-Huxley model for the neuron (which is generally thought to be a more realistics model) shows a decrease of the CV with firing rate. This is highly relevant for interpreting the , , and rhythms in EEG and MEG signals in neuroImaging studies, since these signals can exist thanks to synchronized neuronal activity and since a large variability will hamper synchronous firing.
We will discuss various models for neuronal dynamics and what mechanisms may be responsible for the attention-related modulation of synchronous neuronal activity in action and perception. We will show how a large common input to neurons causes rather small covariances in neuronal firing between two neurons. Therefore, common input is not well coded in the output of a few neurons. However, common input is well reflected in multi-unit output (about 100 neurons). Therefore, the correlation between neuronal firing is a bad predictor of common oscillatory input. This has large implications for interpreting EEG and MEG signals in neuro-imaging in terms of neuronal activity.
Neuroinformatics and cognition: coding and off-line reactivation of representations by neuronal ensembles
C.M.A. Pennartz, University of Amsterdam
The growing field of Neuroinformatics, or Computational Neuroscience, is currently evolving around the development of increasingly realistic models of neural networks, and of sophisticated tools to analyze massively acquired datastreams and get a better grip on neural coding problems. Taking the problem of memory consolidation by neural assemblies as a lead, this talk will focus on the latter development. Learning and memory processes can be expressed at the single-cell level in the brain, but are nevertheless thought to be mediated by large populations of neurons, either within a single brain structure or distributed across multiple areas. This essentially distributed nature of information processing and storage necessitates the development and application of instruments having the capacity to monitor the activity of many neurons simultaneously. Monitoring many-neuron spike trains requires a high temporal resolution and a solution to the spike-sorting problem (viz., how to assign similar spikes to one neuron and dissimilar spikes to different neurons). To a great extent these measuring constraints have been met by the Tetrode-Array Recording Technique, which currently allows us to record 10-70 discriminated neurons simultaneously in brain areas of task-performing rats - ranging from the hippocampus and orbitofrontal cortex to the ventral striatum. Recording ensembles in the latter structure, which is involved in the invigoration of behavioral output and in drug addiction, we show that neurons exhibit 'replay' of firing patterns typical of a behavioral experience during a sleep period that follows this experience. This replay, or reactivation, is emerging as an important model of spontaneous, off-line memory retrieval and possibly consolidation. The various computational steps in gathering the evidence for replay will be laid out, the intriguing properties of replay in the ventral striatum will be contrasted with the hippocampus, and a hypothesis will be presented on the orchestration of reactivation in distributed brain areas by the hippocampus.