GEARNet: grammatical evolution with artificial regulatory networks

The Central Dogma of Biology states that genes made proteins that made us. This principle has been revised in order to incorporate the role played by a multitude of regulatory mechanisms that are fundamental in both the processes of inheritance and development. Evolutionary Computation algorithms are inspired by the theories of evolution and development, but most of the computational models proposed so far rely on a simple genotype to phenotype mapping. During the last years some researchers advocate the need to explore computationally the new biological understanding and have proposed different gene expression models to be incorporated in the algorithms.Two examples are the Artificial Regulatory Network (ARN) model, first proposed by Wolfgang Banzhaf, and the Grammatical Evolution (GE) model, introduced by Michael O'Neill and Conor Ryan. In this paper, we show how a modified version of the ARN can be combined with the GE approach, in the context of automatic program generation. More precisely, we rely on the ARN to control the gene expression process ending in an ordered set of proteins, and on the GE to build, guided by a grammar, a computational structure from that set. As a proof of concept we apply the hybrid model to two benchmark problems and show that it is effective in solving them.