Methods for computational neuroscience.

Preface Methods for computational neuroscience Computational neuroscience, and more generally theoretical neuroscience, has traditionally been looked at with some scepticism by the experimental neuroscience community. One of the reasons for this scepticism was that models were often disconnected from reality, describing situations that were too idealized compared to the actual biological complexity. However, this situation changed following the emergence of computational studies that were tightly based on experimental data, as exemplified by the Hodgkin and Huxley model of the action potential. Models began to be considered as useful tools to understand physiological recordings. Since then, computational neuroscience has experienced a tremendous growth, and more rarely triggers scepticism. Computational methods now not only aim at explaining or predicting experimental observations, but they also provide tools to manipulate and analyze experimental data. They may even interact directly with living neurons, a fact that may have been unimaginable a few decades ago. The goal of this Special Issue on " Methods for Computational Neuroscience " is precisely to overview such computational methods that are directly applicable to experimental data. The first type of application is one that allows models to interact directly with living neurons. This so-called " dynamic-clamp " or " conductance-injection " technique consists of injecting – via the intracellular electrode – conductances in the recorded neuron. Because the current injected in the neuron necessarily depends on the instantaneous value of the membrane potential, a real-time interaction between the computer-generated conduc-tances and the living neuron is necessary. Several papers in this issue directly or indirectly deal with these issues. Robinson describes a new DSP-based system to perform " conductance-injection " experiments, which is programmable and enables a wide range of applications. Bettencourt and co-workers compare real and simulated " dynamic-clamp " experiments, and examine artefacts that arise from this technique and possible ways to correct them. Hughes and co-workers introduce a new simulation system called NeuReal which is capable of simulating artificial dendrites and creating hybrid networks with real and simulated neurons, as well as other applications. Piwkowska and co-workers describe a number of methods to analyze intracellular recordings and extract conductances. They use a " dynamic-clamp " system based on the NEURON simulator to test these methods in controlled situations. These papers depict some of the latest developments in the " dynamic-clamp " technique, which remains one of the closest type of possible interactions between models and experiments. Another theme …