(692a) Genome Alignment and Pair-Wise Comparison of Microbial Metabolic Networks Identifies Novel Potential Drug Targets

To date, genome-scale metabolic reconstructions have been developed for over 100 prokaryotic organisms, and this number is rapidly increasing. These reconstructions serve as the foundation for constraint-based models which can be used to study cellular behavior. While a variety of methods have been developed to understand and improve cellular phenotypes, these constraint-based methods are generally only applied to models of single organisms. Recent studies have compared multiple organisms via their metabolic networks, by mapping metabolic compounds and reactions across the networks and then looking at differences and similarities in reaction and gene content (1, 2). This reaction comparison approach can be time-consuming when different nomenclatures or abbreviations are used to describe metabolites and reactions. Additionally, while a reaction alignment approach can identify network differences, it does not identify the effect of these differences on the functional states of the network. For example, one model may include a reaction as reversible while another model includes it as irreversible, but the impact of this directionality difference is generally not known. Furthermore, differences at the gene level, such as the presence of isozymes or multi-subunit enzymes, will not be identified by comparing reaction contents.

To overcome these weaknesses, we have developed a mixed integer programming approach to identify the functional differences between organisms by aligning models at the gene level. The approach first defines a set of orthologous genes based on genome sequence, and then identifies conditions under which genetic differences give rise to differences in metabolic capabilities. Because a critical issue in this analysis is the assignment of genetic orthologs, we also assessed the accuracy of existing ortholog prediction tools with respect to model annotations, and identified strengths and weaknesses of a number of these tools. We applied the algorithm to existing models of two human pathogens, Mycobacterium tuberculosis and Staphylococcus aureus, in order to explore differences in pathogenicity and drug resistance based on differences in reaction and gene content. By seeking genetic perturbation strategies (i.e. ortholog deletions) that are lethal in only one organism, we were able to identify differences in their metabolic networks (e.g. gene and reaction differences) and network capabilities (e.g. unique metabolic transformations) which point to unique metabolic functions as possible targets for new antimicrobials.  This general gene-centric mixed integer programming approach allows for a rapid comparison of metabolic models based on genomic data without requiring a detailed alignment of metabolic networks.

References

(1) Thiele, et al. “A community effort towards a knowledge-base and mathematical model of the human pathogen Salmonella Typhimurium LT2.” BMC Systems Biology 2011. 5(8).

(2) Oberhardt, et al. “Reconciliation of Genome-Scale Metabolic Reconstructions for Comparative Systems Analysis.” PLoS Computational Biology 2011. 7(3):e1001116.