Characterizing and Ranking Computed Metabolic Engineering Strategies
Conference on Constraint-Based Reconstruction and Analysis (COBRA)
2018
5th Conference on Constraint-Based Reconstruction and Analysis (COBRA 2018)
Poster Session
Poster Session
Sunday, October 14, 2018 - 6:00pm to 7:00pm
Computational strain design, i.e. the calculation of metabolic intervention strategies from a mathematical model, is a key component of an integrated metabolic engineering approach. A broad range of methods has been developed for this task, including bilevel optimization approaches (e.g.,OptKnock and RobustKnock) and the framework of Minimal Cut Sets (MCSs). Most of these algorithms enforce the coupling between growth and the synthesis of the desired product. Some of them, such as the MCSEnumerator, may return a large pool of thousands of possible intervention strategies from which the most suitable strategy can then be selected.
This work focuses on how to characterize and rank, in a meaningful way, a given pool of intervention strategies calculated for growth-coupled product synthesis. Some criteria are straightforward and can be assessed easily, for example, the number of necessary reaction or gene cuts, the maximal growth rate and the guaranteed (minimum) product yield. Additionally, we present more extensive criteria that are worth considering when picking the ‘best’ intervention strategy. Among others we investigate the existence of metabolites that may disrupt the growth coupling when accumulated or secreted and check whether the interventions interrupt pathways at their origin or (less preferred) at later steps. We also assess the maximum thermodynamic driving force of the pathway(s) favored by the intervention strategy and take into account whether a specific intervention strategy relies on flux re-routing within the central metabolism or on other pathways with possibly minor throughputs. Furthermore, strategies that have significant overlap with alternative solutions are also ranked higher because they provide flexibility in implementation.
We applied the presented ranking approach to several sets of intervention strategies for growth-coupled synthesis of native and heterologous products computed from the genome-scale E. coli model iJO1366 using the MCSEnumerator algorithm.
This work focuses on how to characterize and rank, in a meaningful way, a given pool of intervention strategies calculated for growth-coupled product synthesis. Some criteria are straightforward and can be assessed easily, for example, the number of necessary reaction or gene cuts, the maximal growth rate and the guaranteed (minimum) product yield. Additionally, we present more extensive criteria that are worth considering when picking the ‘best’ intervention strategy. Among others we investigate the existence of metabolites that may disrupt the growth coupling when accumulated or secreted and check whether the interventions interrupt pathways at their origin or (less preferred) at later steps. We also assess the maximum thermodynamic driving force of the pathway(s) favored by the intervention strategy and take into account whether a specific intervention strategy relies on flux re-routing within the central metabolism or on other pathways with possibly minor throughputs. Furthermore, strategies that have significant overlap with alternative solutions are also ranked higher because they provide flexibility in implementation.
We applied the presented ranking approach to several sets of intervention strategies for growth-coupled synthesis of native and heterologous products computed from the genome-scale E. coli model iJO1366 using the MCSEnumerator algorithm.