First published 2 July 2012
Evolved Modular Epistasis in Artificial Organisms
Sergi Valverde, Ricard V. Solé, Santiago F. Elena
How does complexity evolve in artificial and natural systems? A central concept within genetic systems is epistasis, namely the modulation of the effects of a given gene by one or several other genes. Epistasis is known to have an impact on many features of organisms, from recombination and sex to the ruggedness of the underlying fitness landscapes. However, the multi-scale nature of evolution and organisms makes often difficult to properly characterize epistatic interactions. Here we study the hierarchical organization of epistatic interactions between machine instructions in evolved digital organisms. We present a new quantitative approach to discover epistatic interactions that is able to capture the presence and role of groups of epistatic modules. Therefore, it thus takes into account the intrinsic nested nature of individual complexity. We found evidences of modular epistasis in avidians, with some modules having a tendency toward antagonistic epistasis while others show the opposite epistatic sign. We also found that this modular organization was positively correlated to organismal robustness.