Differential co-expression in DNA microarray data and it's application to animal health

Many of the interactions between the host and pathogen during infection and disease occur at the molecular or cellular level. DNA, RNA and protein molecules, from both host and pathogen, all interact with one another to create the disease and the host immune response, both of which are interesting to investigate. Techniques developed over the last ten years now enable us to measure the quantities of these molecules that are present in any given biological sample. By carefully designing experiments involving both host and pathogen, both before and after infection, we can gather data on the differences between these states in terms of the biologically important molecules which are present. Combining data from these various experiments in order to infer how these molecules interact is a large challenge which will require skills in computing, mathematics and statistics. Recent publications, including one by the author of this proposal, have shown that analysing when and where these moelcules interact (termed co-expression) leads to new information regarding the mechanism of disease and the host immune response. We intend to develop and apply these techniques to data gathered at the institute for animal health on important farm animal diseases, in an attempt to understand infection, disease and the host immune response.

Many of the interactions between the host and pathogen during infection and disease occur at the molecular or cellular level. The wealth of information provided by post-genomic studies makes them ideal for investigating complex systems. Microarrays have become a standard tool for the exploration of global gene expression changes at many levels, including different cells, tissues, developmental stages, disease states or samples subjected to appropriate treatments. Recent research on the analysis of microarray data has concentrated on the analysis of co-expression. This approach recognizes that genes do not exist in isolation, and that it may be disruptions in the relationships between genes that cause and/or result from biological phenotypes, not simply changes in their level of expression. The author of this proposal has published the most recent work in this area. CoXpress is a package for R that detects differential co-expression by examining groups of genes that are highly correlated in one data set but no more correlated than can be expected by chance in a second data set. The purpose of this proposal is to develop the CoXpress package into a resource that will be valuable to researchers in animal health and many other areas. We will implement improved methods of defining groups of co-expressed genes, as well as adding network construction methods. We will go on to use CoXpress to develop a database of co-expressed genes, limited to datasets interesting to animal health in the first instance. We will investigate the possibility of using co-expression information for class prediction by applying random forest and genetic algorithm approaches to co-expression information. We will use CoXpress to re-analyse datasets produced by the IAH in the context of co-expression, and use the results to inform the functional annotation of genes in both hosts and pathogens of interest to animal health.