Modelling the Brain

Introduction:
Scientists at the Krasnow Institute in the US are interested in the neurological basis of consciousness. The approach they are using in their research is to model entire brains using biologically realistic computer simulations of neurons and neural networks.

The research group, led by Giorgio Ascoli and financed by the Human Brain Project, has developed two sets of simulation software. One, called L-Neuron, is able to realistically simulate single neurons. The second, called Abor Vitae, simulates networks containing large numbers of neurons and their interconnecting synapses. Abor Vitae has already been used to simulate a small part of a rat brain, namely the hippocampus - an area involved in associative memory. It is hoped that in coming years computers will soon become powerful enough to model an entire brain.

Simulated Neurons:
The L-Neuron software creates virtual neurons which are anatomically indistinguishable from their biological counterparts. Using simple and recursive mathematical algorithms, it can create virtual neurons whose dendrites grow, and branch, like real neurons.

By changing a few parameters, different types of virtual neurons can be grown. Their morphology closely resembles that of the various kinds of real neurons, such as pyramidal, granule, or stellate cells. Because each cell is created with a mathematical algorithm, rather than a detailed plan of the position and diameter of its every dendrite, hundreds and thousands of neural tracings can be described with just a small amount of code. L-Neuron is a DOS program which will run on Pentium PC's, and can generate thousands of neurons in seconds.

Morphology and Behaviour:
Neurons come in a large variety of different shapes. It is believed that this morphological variation plays a major role in defining the computational behaviour of neurons.

A large part of the research at Krasnow is into how the geometry and topology of neurons effects their electrical behaviour. Using large simulations, parameters such as ionic concentrations and axonal conductances, etc, are kept the same, whilst dendrite lengths, diameters, taper rates, etc, are changed. It is then studied how these changes in morphology affect the behaviour, such as neuron spiking rates, axon growth, etc.

Simulated Hippocampus:
The Abor Vitae software uses mathematical algorithms similar to those used in L-Neuron, but on the network level. As well as dentritic growth and branching, the algorithms are also able to describe axonal navigation and synaptic connectivity.

Already the team have built a small model of a slice of rat hippocampus. They are now working on a larger hippocampal slice. As the hippocampus is lamela in structure, it is hoped that the whole structure can be assembled by stacking many of these slices together.

ArborVitae currently runs on Silicon Graphics machines, and can grow a complete interconnected million-compartment network in less than ten minutes.

Links:
Computational Neuroanatomy Group: www.krasnow.gmu.edu/ascoli/CNG