A Microbial Screening in Silico Method for the Fitness Step Evaluation in Evolutionary Algorithms

Abstract

One of the most delicate stages of an evolutionary algorithm is the evaluation of the goodness of the solutions by some procedure providing a fitness value. However, although there are general rules, it is not always easy to find an appropriate evaluation function for a given problem. In the biological realm, today, there is a variety of experimental methods under the name of microbial screening to identify and select bacteria from their traits, as well as to obtain their fitness. In this paper, we show how given an optimization problem, a colony of synthetic bacteria or bacterial agents is able to evaluate the fitness of candidate solutions by building an evaluation function. The evaluation function is obtained simulating, in silico, a bacterial colony conducting the laboratory methods used in microbiology, biotechnology and synthetic biology to measure microbial fitness. Once the evaluation function is built, it is included in the code of the genetic algorithm as part of the fitness routine. The practical use of this approach is illustrated in two classic optimization problems. In silico routines have been programmed in Gro, a cell programming language oriented to synthetic biology, and can easily be customized to many other optimization problems.

A Framework for Implementing Metaheuristic Algorithms Using Intercellular Communication

“…Integration with automated techniques for evaluating fitness functions ( Gargantilla Becerra and Lahoz-Beltra, 2020 ) is an immediate expansion to the workflow which can lead to further automation in the definition of the algorithms to generate. Current research is also being invested into relating different AI algorithms such as Neural Networks, Reinforced Learning (Q-Learning) and other MH, such as Ant Colony Optimization, to our framework.…”

“…Furthermore, many computational methods can be translated to Synthetic Biology constructs that emulate their operation, expanding the array of techniques that can be applied on a same problem and improving automation over Directed Evolution. Moreover, cell colonies of synthetic bacteria have been successfully used to build the evaluation function of an EA to evaluate the fitness of candidate solutions ( Gargantilla Becerra and Lahoz-Beltra, 2020 ). It is in this spirit that we study Metaheuristic procedures (MH) ( Glover and Kochenberger, 2006 ; Talbi, 2009 ; Sörensen, 2015 ), a larger class of procedures that contain EAs.…”