Table 1.

Redundancy, phenotype bias and components in Polyomino, RNA and HP GP maps. Comparing the number of phenotypes (NP) to the number of genotypes (NG) for each GP map highlights large-scale redundancy present. Phenotype bias is demonstrated in each map with the measure of the fraction of phenotypes that covers 95% of the genotypes (Ncov is the number of phenotypes that covers the 95% of genotypes). In all cases the fraction of phenotypes is significantly smaller than the fraction of genotypes being covered, indicating the presence of a strong phenotype bias. The final column is the total number of genotype components (NC) in each GP map. In all cases (non-computable values left out), the number of components is larger than the number of phenotypes, indicating phenotypes tend to be spread out over multiple disconnected components. RNA data for L = 12 were computed from the Vienna package [22] and taken from [41] for L = 15, 20. The value of Np for the Polyomino S4,16 GP map is approximate as it is found from large-scale sampling of the GP map over multiple runs of the algorithm presented in [42]. All other Polyomino results were found by exhaustive enumeration. The HP results were calculated from the data made available by Irbäck & Troein [36]. Non-deterministic phenotypes and the trivial structure in RNA are excluded from the statistics.

GP mapNGNPNcov/NP (%)NC
polyomino S2,81.7 × 10713541347
polyomino S3,86.9 × 101014716
polyomino S4,161.8 × 1019∼22373
RNA, L = 121.7 × 1075747645
RNA, L = 151.1 × 1094312312 526
RNA, L = 201.1 × 101211 21810
HP, L = 253.4 × 107107 33668148 254