Multi-objective memetic meta-heuristic algorithm for encoding the same protein with multiple genes

Abstract

An important goal in synthetic biology is to maximize the expression levels of proteins. For this purpose, multiple genes encoding the same protein can be integrated into the host genome. However, this approach is affected by two key issues. Firstly, codons with better adaptation indexes should be used, since some synonymous codons are better adapted than others. Secondly, the multiple protein-coding sequences should be as different as possible to avoid the loss of gene copies due to homologous recombination. Therefore, this task shows strict biological requirements that make it difficult to tackle. In this work, we design and implement a computational intelligence approach to address this problem, the Multi-Objective Shuffled Frog Leaping Algorithm (MOSFLA). This method combines the optimization capabilities provided by parallel searches, multiple operators, and memetic strategies to tackle problems with difficult solution quality requirements. Several alternatives have been comparatively analyzed, including MOSFLA variants with three objectives as in other approaches from the literature and also variants with only two objectives. Experiments on nine real-world protein datasets give account of the improved, statistically significant performance achieved over the related work, attending to different quality metrics, confirming that our proposal satisfactorily deals with the complex nature of the problem.