It has traditionally been believed that information in the brain is encoded either in averaged neuronal activity levels (rate coding), or in the timing of individual action potential pulses (temporal coding). Our recent simulation results and experimental data suggest that the cerebellum, the part of the brain implicated in motor control and motor leaning, might use a different coding strategy based on the duration of silent periods in neuronal activity. So far we have studied this novel type of coding with a detailed model of a single cerebellar Purkinje cell, the major type of neuron in the cerebellar cortex. We suggest the following projects, which build on our previous work:

1. The effect of non-specific synaptic plasticity

   Most models of associative memory assume that neuronal activity patters are stored in networks by selectively changing the weights of synapses between active input and output units. More recently, it has become clear that some networks of neurons in the brain also include non-specific synaptic plasticity that spreads to neighbouring inactive synapses. The non-specific synaptic plasticity in Purkinje cells has been well characterised experimentally and has been implicated in the coordination of cerebellar modules. The aim of this project is to study the effect of non-specific synaptic plasticity on associative memory, both in single Purkinje cells and in the cerebellar network.

2. Decoding of silent periods by the cerebellar output layer

   The Purkinje cells that were the main subject of our previous research project to neurons in the deep cerebellar nuclei (DCN), which provide the only output from the cerebellum and are therefore important for motor control. The aim of this project is to study the decoding of silent periods in Purkinje cell activity by DCN neurons. The computer simulations can initially be based on existing models of cerebellar neurons, including a detailed model of DCN neuron that was developed in collaboration with Dieter Jaeger at Emory University. In later stages of the project, the detailed neuronal models could be optimised and simplified for easier use in a large network.

3. Associative memory in a network model of the cerebellar cortex

   Until now we have studied pattern recognition in a single Purkinje cell. However, the behaviour of Purkinje cells is determined by the network of neurons that provide their direct or indirect inputs. The aim of this project is to construct and investigate a 3D model of the cerebellar cortex. The network model can then be used to study associative memory and other theories of cerebellar function relating to different forms of coding. A potential extension of this project is to automatically generate variable neuronal morphologies and study the effect of variability of network activity.

Although these projects require an interest in the underlying neurobiology, no previous biological knowledge is required. The simulations can be implemented using neuronal simulation software such as GENESIS and NEURON. Moreover, the network construction and visualisation software neuroConstruct, which is being developed in collaboration with Padraig Gleeson and Angus Silver at UCL, can be used for the development of networks of neurons in 3D.

To discuss these project please contact Volker Steuber.

The Human Genome project and similar projects for other species have succeeded in sequencing many genomes for different species. However the next major task involves analysing the complicated interactions between the components parts of these sequences, namely the genes and their regulatory regions.

We are currently developing techniques for improving the accuracy of computational predictions for regulatory signals in genomic sequences. We would now like to extend our work, originally developed using yeast as a model organism, to other organisms exhibiting more complex regulatory organisation. We would also like to extend our analysis to include new algorithms and techniques and to investigate any relationship between different algorithmic strategies and the underlying structure of the regulatory DNA. This project would suit a student with either a biological background or a computational background with an interest in biology.

To discuss this project please contact Mark Robinson, or Rod Adams.

Members of the Biocomputation Research Group are exploring the structure and function of molecular networks that control the behaviour of cells. Knowing that many diseases stem from, or result in changes in the dynamics of these networks, we want to understand how biomolecules interact to regulate processes such as gene expression and motility in living cells. Together with experimental scientists in the UK and abroad, we try to identify important components and regulatory fluxes by reconstructing the network in silico and analyzing the behaviour when subject to perturbation. In collaboration with the members of the Computer Science Department, we do computational research on the function of neurons, and on the functional organization of genomes.