Methods for the refinement of genome-scale metabolic networks

Abstract

More accurate metabolic networks of pathogens and parasites are required to support the identification of important enzymes or transporters that could be potential targets for new drugs. The overall aim of this thesis is to contribute towards a new level of quality for metabolic network reconstruction, through the application of several different approaches. After building a draft metabolic network using an automated method, a large amount of manual curation effort is still necessary before an accurate model can be reached. PathwayBooster, a standalone software package, which I developed in Python, supports the first steps of model curation, providing easy access to enzymatic function information and a visual pathway display to enable the rapid identification of inaccuracies in the model. A major current problem in model refinement is the identification of genes encoding enzymes which are believed to be present but cannot be found using standard methods. Current searches for enzymes are mainly based on strong sequence similarity to proteins of known function, although in some cases it may be appropriate to consider more distant relatives as candidates for filling these pathway holes. With this objective in mind, a protocol was devised to search a proteome for superfamily relatives of a given enzymatic function, returning candidate enzymes to perform this function. Another, related approach tackles the problem of misannotation errors in public gene databases and their influence on metabolic models through the propagation of erroneous annotations. I show that the topological properties of metabolic networks contains useful information about annotation quality and can therefore play a role in methods for gene function assignment. An evolutionary perspective into functional changes within homologous domains opens up the possibility of integrating information from multiple genomes to support the reconstruction of metabolic models. I have therefore developed a methodology to predict functional change within a gene superfamily phylogeny.