Principal Investigator Steve J Redman Project r29

Division of Neuroscience, Machine CM

John Curtin School of Medical Research

Co-Investigator Allan Coop

Division of Neuroscience, John Curtin School of Medical Research

Hippocampal Local Circuits

This research aims to use biologically realistic neural networks to help understand the activity of complex neuronal circuitry. We tackle complex parts of the nervous system where we have expert knowledge, such as the hippocampus, to attempt to simulate various features of their operation.

What are the basic questions addressed?

What is the role of the inhibitory interneurones in the integrated behaviour of the hippocampus?

What are the results to date and the future of the work?

A model of a circuit based on data drawn from area CA1 of the rat and guinea-pig hippocampus has been developed. The circuit contained 27 excitatory neurones and one each of three different types of inhibitory interneurone to give a total of 30 cells. Each cell was modelled as a single isopotential compartment with its membrane represented by an RC circuit. This approach reduced the number of free parameters and increased the computational feasibility of simulations. Each neurone could generate a fast excitatory and a fast inhibitory synaptic potential in response to synaptic activation. The excitatory neurones and two of the three inhibitory interneurones could also generate a slow inhibitory synaptic potential. The cells were tuned to give a biologically realistic response to just a single suprathreshold stimulus. The circuit formed by the synaptic connections between the cells was considered to be biologically realistic. This circuit had no connections between excitatory neurones and no connection between the excitatory neurones and the interneurone that generated slow inhibitory postsynaptic potentials. Simulations were used to determine the effect that altering the number of afferent fibres that converged onto each neurone, and changes in the magnitude of synaptic weights, had on the discharge of each type of cell. The effect on the discharge of cells due to the removal of a particular type of synapse, or the removal of one or more interneurones was also examined.

Simulations indicated that discharge of the excitatory cells was not under the control of any one interneurone. Further, that the number of afferent fibres that converged onto each type of interneurone determined whether an increase in inhibitory drive acted to increase or reduce discharge of the excitatory neurones. It was found that for a particular set of synaptic weights, the frequency of stimulation could determine which synaptic connection or type of interneurone acted to control discharge of the excitatory cells. Finally, the removal of any one type of interneurone had little effect on discharge of the excitatory neurones, but the removal of any two types of interneurone acted to reduce the discharge of excitatory cells.

What computational techniques are used and why is a supercomputer required?

The cells in the neural network were modelled by first order partial differential equations using an exponential integration scheme. The large amount of computation and the memory required to find a suitable set of parameter values for each cell type, and to run the final simulations, suited the CM5.