Genetic Correlations Greatly Increase Mutational Robustness and Can Both Reduce and Enhance Evolvability
Fig 4
Phenotype mutation probabilities scale with global frequency.
We present results for the three GP maps: A) RNA20, B) S3,8 and C) HP5x5. We plot the relationship between ϕqp (circles) and fq for the largest non-deleterious phenotype p in S3,8 and HP5x5, and for the second largest in RNA20 (not the largest due to computational expense). We see in each case a strong positive correlation (p-value ≪ 0.05 in all cases), very similar to the expectation for the null model (not shown here, but for which the correlation is exact to within statistical fluctuations, see ref. [14] and S1 Text). Spearman rank correlation coefficients are shown in the top-left of each plot. Differences from ϕqp = fq are relatively small compared to the overall range of variation, except for sets of phenotypes that are not connected at all, which typically arise due to biophysical constraints. These are shown as downward triangles along the lower horizontal dotted line which represents ϕqp = 0. For each plot, the upward triangle indicates ϕpp = ρp, the phenotype robustness, which is always over-represented (ρp ≫ fp) due to neutral correlations.