Canalization of the Genotype-Phenotype Map: A Coding Theoretic Perspective

In the evolutionary biology literature, canalization refers to the mechanism by which variability is reduced at the phenotypical level. It was first observed by Waddington in 1942 and has been the focus of recent research interest. Here, we attempt to build a bridge between evolutionary biology and information theory by offering a fresh coding theoretic perspective of canalization. In particular, we focus on evolution in rugged multi-peak fitness landscapes and argue that canalization can efficiently smooth-out the roughness of the landscape, and hence, facilitate easier transitions from lower to higher peaks. Inspired by coding theory, we view the landscape peaks as codewords and allow our population to learn the underlying codebook and decision regions through evolution. Interestingly, our model sheds more light on the interplay between robustness, neutrality, and evolvability. Finally, we support our claims by numerical results generated through Monte-Carlo simulations in polynomial fitness landscapes.