Elements of Computational Systems Biology

16

VALIDATION ISSUES IN REGULATORY MODULE DISCOVERY

Alok Mishra and Duncan GilliesDepartment of Computing, Imperial College London, London, UK

16.1 INTRODUCTION

Over the last decade, many new techniques have emerged for measuring the functional activity of genes. These include measurements of individual genes through their transcriptional activity, the interaction of their products, or by in vitro experiments with synthesized DNA. For each measurement technique a different set of data is created, each containing information on the functioning of genes, but under differing conditions and with different degrees of experimental error. At the same time, summary information of accepted gene behavior is being collected in the form of the gene ontology database and annotation terms for individual genes. These resources represent accumulated knowledge rather than individual experimental data. Current research is being undertaken to investigate ways in which these rich but diverse sources of information about gene behavior can be combined to provide a more accurate interpretation of experimental work. The goals are two-fold. Firstly, fusing data from diverse sources can be used to stabilize the results from individual experiments. For example, microarray experiments have been generally found to produce data of high variance, and therefore require some form of regularization before results can be interpreted. Secondly, when the data are processed to infer some higher organization among genes, ...