Hotel Estoril Eden, Monte Estoril,
5-8 October 2005



NextText Box: Participants
Text Box: Programme

Protein Expression Data / Genetic Networks

Marco Grzegorczyk and Klaus Jung
Department of Statistics, University of Dortmund, Germany

On this poster we concentrate on two different topics. First the analysis of protein expression data from two-dimensional gel electrophoresis. Second the determination of interacting genes using Bayesian networks.

Difference gel electrophoresis (DIGE) is an improvement of two-dimensional gel electrophoresis which allows the user to put up to three different mixtures on the same gel. This technique reduces the gel-to-gel variation. We show how to calibrate and normalize this data. We give a mixed linear model for finding time-treatment interactions in longitudinal DIGE data. A big problem in gel data is the existence of many missing values. Therefore, we propose a method for the estimation of the missing data points using the k nearest neighbour method.

Afterwards, we give a brief introduction to the theory of Bayesian Networks (BNs) and present results obtained by applying Gaussian and multinomial BNs with different MCMC and Bootstrap sampling schemes to synthetic, but biologically realistic, gene expression data. In addition, we compare these results with those obtained by less computational expensive network approaches like Gaussian Graphical Networks and the mutual information based Relevance Networks methodology.