The Aevol model. a) Individuals are distributed on a grid. At each generation, the whole population replicates according to a Wright–Fisher replication model, in which selection operates locally within a $3×3$ neighborhood. b) Each grid cell contains a single organism described by its genome. Genomes are decoded through a genotype-to-phenotype map with 4 main steps (transcription, translation, computation of protein functions, and computation of the phenotype). Here, for illustration purposes, a random gene and the corresponding mRNA are colored in red. The red triangle represents the function of this gene in the mathematical world of the model. The phenotypic function is calculated by summing all protein functions. The phenotype is then compared with a predefined target (in green) to compute the fitness. The individual presented here has evolved in the model during $500,000$ generations. c) Individuals may undergo mutations during replication. Two example mutations are shown: A small insertion (top) and a large deletion (bottom). Top: A 1 bp insertion occurs within a gene. It causes a frameshift, creating a premature stop codon. The ancestral function of the gene is lost (dashed triangle) and the truncated protein has a deleterious effect (red triangle). This leads to a greater divergence between the phenotype and the target (orange area on the phenotype). Bottom: The deletion removes 5 genes. The functions of 2 of them can be seen in the box (dotted triangles). This results in a large discrepancy between the phenotype and the target (orange area on the phenotype).