In bioinformatics, metabolism is usually modeled by directed graphs
where nodes represent reactions and edges represent substrate/product
exchanges between them. Subsets of these networks are often used to
represent collections of reactions involved in a same biological
process, also called metabolic pathways. Enzymes recruited in these
pathways have often the capacity of catalyzing one or several types of
chemical reactions on more or less different substrates. The aim of my
work is to explore the metabolism in search of pathways of chemical
transformations that were conserved during evolution. To detect these
conserved sequences of chemical transformations, I propose a graph
formalism where the metabolic network is transformed to a network of
reaction signatures, merging reactions that share the same type of
chemical transformation in a same node. This representation allows the
definition of several conservation metrics depending on the number of
reactions by chemical transformation type, the number of enzymes
associated to each type, and the topological importance of nodes and
edges. Using these metrics, scores are computed for known and enumerated
pathways to evaluate their degree of conservation. The detection of
these chemical units, also named reaction modules, will reveal new
metabolic pathways for which reactions and enzymes remain unknown.