Dennis Egen1, Desmond S Lun. 1. Center for Computational and Integrative Biology and Department of Computer Science, Rutgers University, Camden, NJ 08102, USA.
Abstract
MOTIVATION: Computer-aided genetic design is a promising approach to a core problem of metabolic engineering-that of identifying genetic manipulation strategies that result in engineered strains with favorable product accumulation. This approach has proved to be effective for organisms including Escherichia coli and Saccharomyces cerevisiae, allowing for rapid, rational design of engineered strains. Finding optimal genetic manipulation strategies, however, is a complex computational problem in which running time grows exponentially with the number of manipulations (i.e. knockouts, knock-ins or regulation changes) in the strategy. Thus, computer-aided gene identification has to date been limited in the complexity or optimality of the strategies it finds or in the size and level of detail of the metabolic networks under consideration. RESULTS: Here, we present an efficient computational solution to the gene identification problem. Our approach significantly outperforms previous approaches--in seconds or minutes, we find strategies that previously required running times of days or more. AVAILABILITY AND IMPLEMENTATION: GDBB is implemented using MATLAB and is freely available for non-profit use at http://crab.rutgers.edu/~dslun/gdbb.
MOTIVATION: Computer-aided genetic design is a promising approach to a core problem of metabolic engineering-that of identifying genetic manipulation strategies that result in engineered strains with favorable product accumulation. This approach has proved to be effective for organisms including Escherichia coli and Saccharomyces cerevisiae, allowing for rapid, rational design of engineered strains. Finding optimal genetic manipulation strategies, however, is a complex computational problem in which running time grows exponentially with the number of manipulations (i.e. knockouts, knock-ins or regulation changes) in the strategy. Thus, computer-aided gene identification has to date been limited in the complexity or optimality of the strategies it finds or in the size and level of detail of the metabolic networks under consideration. RESULTS: Here, we present an efficient computational solution to the gene identification problem. Our approach significantly outperforms previous approaches--in seconds or minutes, we find strategies that previously required running times of days or more. AVAILABILITY AND IMPLEMENTATION: GDBB is implemented using MATLAB and is freely available for non-profit use at http://crab.rutgers.edu/~dslun/gdbb.