In silico modelling of directed evolution: Implications for experimental design and stepwise evolution |
| |
Authors: | David C. Wedge William Rowe Douglas B. Kell Joshua Knowles |
| |
Affiliation: | a Manchester Interdisciplinary Biocentre, University of Manchester, 131 Princess Street, Manchester, M1 7ND, UK b School of Computer Science, University of Manchester, Kilburn Building, Oxford Road, Manchester, M13 9PL, UK c School of Chemistry, University of Manchester, Oxford Road, Manchester, M13, UK |
| |
Abstract: | We model the process of directed evolution (DE) in silico using genetic algorithms. Making use of the NK fitness landscape model, we analyse the effects of mutation rate, crossover and selection pressure on the performance of DE. A range of values of K, the epistatic interaction of the landscape, are considered, and high- and low-throughput modes of evolution are compared. Our findings suggest that for runs of or around ten generations’ duration—as is typical in DE—there is little difference between the way in which DE needs to be configured in the high- and low-throughput regimes, nor across different degrees of landscape epistasis. In all cases, a high selection pressure (but not an extreme one) combined with a moderately high mutation rate works best, while crossover provides some benefit but only on the less rugged landscapes. These genetic algorithms were also compared with a “model-based approach” from the literature, which uses sequential fixing of the problem parameters based on fitting a linear model. Overall, we find that purely evolutionary techniques fare better than do model-based approaches across all but the smoothest landscapes. |
| |
Keywords: | Genetic algorithm Fitness landscape NK-landscape Selection pressure Mutation rate |
本文献已被 ScienceDirect 等数据库收录! |
|