Skip to main content Skip to navigation

COMBI Research in Computational Neuroscience

Modelling in Neuroscience

We have a wide range of interests and we have built models from molecular (LTP, LTD, hormone), to neuron (abstract and biophysic), to neuronal networks and to behaviour level (motor control). If it is possible, all models are data driven and we are working very closely with biologists. Here is a list of projects in this domain:


LTP+LTD and Intracellular Dynamics

We are developing models to account for experimentally observed intracellular signalling pathways. Long term goal is to integrate subcellular, cellular, network and behaviour level to study learning and memory.


Hormone Release

Working with Leng G (Edinburgh), Kendrick K. (Cambridge), we are developing models on oxytocin and studying its impact on behaviours.


Single Neuron Dynamics

We have worked in the area for many years: abstract model such as the integrate and fire model and biophysically realistic model. Early work included a biophysical model for the mitral cell and a biophysical model for CGC cell etc. Recent work is on modelling Purkinje cell in cerebellum with a few thousand compartments.


Neuronal Networks

We have developed two types of models: the abstract model and biophysical model. For abstract model, we have developed moment neuronal network to deal with networks of integrate-and-fire models. Current work in this area is being conducted with our Fudan colleagues (Prof. Lu WL, Dr. Liu B. et al.).


Motor Control

If we agree that neurons send out stochastic signals, a natural question is then how we can achieve a precise control using a stochastic signals: a stochastic control task. In the past few years, we have worked on various control problems: mainly from optimal stochastic control point of view.


Virtual Nucleus

Together with our colleagues from the Babraham institute (Dr. Peter Fraser), Cambridge, we are working on developing a platform to simulate and analyze transcription dynamics in a nucleus. We have developed a simple model (which treats each gene as a point) to study the transcription dynamics. We are developing more realistic models (polymer as chromosome) and eventually we hope to have a virtual nucleus.


Granger Causality

Granger causality has become a very powerful technique to reveal network structures from a give set of time series data. In recent years, we have developed many new techniques to apply the Granger causality to neuroscience and molecular biological data.


Community Networks

We have developed novel approaches to tackle high dimensional and dynamical data. In recent years, we are in particular interested in exploring structures such as communities in a give data set. We pioneeringly introduced Young measure to deal with the issue of Lpregularization, where 0<p<1, and proved that the solution is localized and has many nice properties.


Metabolic Circuits and Genetic Circuits

We are working on an interesting issue on two circuits: one for nutrition and the other for genetics. Working with Prof. Jin Li, Prof. Lin Wei, both from Fudan, and others, we are investigating how the lack of folate could cause neuronal tube defect (NTD).