Dr Guiyuan Lei
University of Newcastle
Newcastle upon Tyne, United Kingdom
Research Interests:
Reverse engineering of genetic regulatory netowks is to discover the interactions between genes from the experiment, especially from microarry data. We use both continuous and discrete model to describe the interactions. Gaussian Graphical Model, Ordinary Differential Equation (ODE) and Bayesian Network (BN) are main models we use.
• Monte Carlo and Quasi-Monte Carlo
Monte Carlo and Quasi-Monte Carlo methods are the topics of my Ph.D. thesis. Monte Carlo is widely used in three major subjects: Numerical integration for high-dimensial problem, simulation and optimization. Now I am developing Markov Chain Monte Carlo (MCMC) to search the best Bayesian Network. The continuous Quasi-Monte Carlo optimization method is also a potential method to optimize the parameter (coefficients) of ODE.
• Genetic Programming (including Genetic Algorithms)
Genetic Programming (GP) is the extention of Genetic Algorithms (GA). It can present more complex problems and solve the problem using evolution strategy. You can present the right hand side of ODE as a program tree and let computer automately find the ODE which fits data best. The ability of GP has been demonstrated in the field of computational biology.
Bayesian Network (BN) is a kind of Directed Acyclic Graph (DAG). My work focus on how to search the best BN from the experiment data and how to combine the prior knowledge of biology system to transfer the Mathematical model into physical model.
Another network which is important to represent the gene regulate network is Boolean network. You can study the dynamical property of a system using Boolean network and learn the wonderful universe of biology.
Data mining seems to include too many things. The above topics can be contained in this topic. The other thing which I am interested in is how to search useful information from tremendous amount of source in the internet.
Last modified:
11 October, 2007