Abstract
Cells initiate responses to the external environment and internal state through a complex network of signaling pathways composed of many multipurpose proteins. Errors within these signaling networks play a major role in disease progression, so determining signaling activity is crucial in understanding many diseases, including cancer. The changes in these pathways lead to multiple responses within cells, including induction of transcription, allowing the use of microarray data for interpretation of signaling activity. However, linking transcriptional changes to signaling pathways is complicated by the multipurpose nature of proteins, the overlap of signaling pathways, the presence of routine background transcription of housekeeping genes, and the lack of correlation between transcript and protein levels. In order to recover estimates of signaling from microarray data, several steps are required, including (1) modeling of signaling pathways and their links to transcription factors, (2) analysis of transcription factor and transcription complex binding sites in the genome, (3) use of Bayesian methods to extract overlapping transcriptional signatures, and (4) determination of the appropriate dimensionality for analysis. Here we present an approach using simplistic network models, existing databases of transcription factors, and Bayesian Decomposition to demonstrate the methodology.
Introduction
Many methods have been developed to model biological processes and extract information from transcriptional data, and this review cannot fully cover all such methods. The goal will be instead to provide an overview of the key issues to resolve in estimating signaling changes when working with transcriptional data, as provided by GeneChips and microarrays.
Signaling Networks and Transcription
Signaling networks provide cells with the ability to initiate complex responses to the external environment and internal state.