Relevance Networks: A First Step Toward Finding Genetic Regulatory Networks Within Microarray Data

An increasing number of methodologies are available for finding functional genomic clusters in RNA expression data. In this chapter, we describe a technique, termed relevance networks, that computes comprehensive pairwise measures of similarity for all genes in such a dataset. Associations with high positive or negative measures are saved and displayed in a graph-network-type diagram. Advantages of this method over others include: (1) negative associations (e.g., those from tumor suppressing genes) are shown: (2) disparate data types can be included (i.e., clinical, expression, and phenotypic); and (3) multiple connections are allowed (e.g., a transcription factor may be responsible for regulating the expression of multiple other genes). Java-based software is available for academic use to construct relevance networks, and operation of the software is also explained in this chapter.