Integrated analysis of the gene neighbouring impact on bacterial metabolic networks

Different levels of abstraction are needed to represent a living system. Unfortunately information of different nature is not superposable in an obvious way, but requires a dedicated framework. Because biological abstractions, i.e., genomic or metabolic information, can be easily respresented as graphs, it is intuitive to integrate them into a unique graph, in which one can perform graph analysis for investigating a given biological assumption. This study follows such a philosophy and completes a genome and metabolome combination. In a such integrated framework and as illustration, we applied a graph analysis that automatically investigates impacts of the gene adjacency to predict functional relationships between genes and reactions. Our approach, called SIPPER, creates a weighted graph, in which the weights rely on the given relationship between genes, and computes (alternative) chains of reactions catalysed by genes. This method, as a generalisation of methods already published, can be easily adapted to several biological assumptions, properties or measures. This paper evaluates SIPPER on Escherichia coli. We automatically extract subgraphs, called k-SIPs, and quantify their interest in both genomic and metabolic contexts by showing functional compounds like operons or functional modules. [Includes supplementary material]