Information Technoloy for Gene Network Analysis

点击翻译按钮获取中文摘要

We developed information technology for estimating gene networks accurately from multiple source of genetic information including microarray gene expression data. We also developed an XML format for describing dynamic models of gene networks. The following results are obtained:(1)We developed a computational method for estimating gene networks from microarray data obtained from various perturbations such as gene disruptions, gene overexpressions, drug responses, etc. The method combines the Bayesian network approach with nonparametric regression, where genes are regarded as random variables and the nonparametric regression enables us to capture from linear to nonlinear structures between genes. As a criterion for choosing good networks, we introduced the BNRC (Bayesian network and Nonparametric Regression Criterion) score. Naturally, the sole use of microarray data has limitations on gene network estimation. For improving the biological accuracy of estimated gene networks, we have made … More a general framework by extending this method so that it can employ genome-wide other biological information such as sequence information on promoter regions, protein-protein interactions. Computational experiments were conducted with yeast data and they show that cascades of gene regulations were effectively extracted.(2)The problem of finding an optimal Bayesian network is known computationally intractable for BNRC, BDe, MDL scores. For example, in order to find an optimal Bayesian network of 20 genes from 100 microarray data, the brute force algorithm employing all computing resources in the world even requires the time exceeding the life time of the solar system. Our recent computational challenge has made possible to search and enumerate optimal and suboptimal Bayesian networks in feasible time on supercomputers. Computational experiments with this search algorithm have provided evidences of the biological rationality of our computational strategy We obtained the following scientific knowledge: (i) Optimal Bayesian networks do not necessarily reflect the most accurate biological knowledge. (ii) Gene・gene relations which appear very frequently in optimal and suboptimal networks are biologically very likely. (iii) BNRC score is much better than BDe and MDL scores for gene network estimation. Furthermore, by putting constraints arising from biological knowledge, a faster algorithm is also developed.(3)We designed and developed an XML called CSML Version 1.0 (Cell System ML) for describing and simulating dynamic models of gene networks. Based on this, we developed programs which automatical convert pathway models in KEGG and BioCyc to CSML models so that dynamic models can be constructed. Cell Illustrator is used for modeling and simulation of the models which employs CSML for model description. This framework enables us to develope large-scale metabolic pathway models for simulation. Less