Figure 4.

Deviant codes predicted from SGC represented as an evolutionary tree. Only the codon reassignments found in known deviant codes are shown. The anticodon recognition pattern of existing mitochondrial codes (for examples, see refs. 5 and 23) was used to predict mitochondrial codes. The numbers labeled total indicate the total number of codes with 21 phenotypes. The edges are labeled with the number of predicted codes with a similar manner of improvement, the optimality of the reassignment, and the numbers in parentheses are successive improvement rates in the changeability, μ-robustness, and s-robustness. Note that a positive improvement in the s-robustness means a decrease in ϕ. Hyp code means hypothetical, unknown genetic codes. For example, of the total of 260 possible codes derived from the code CMy that code 21 phenotypes, 1/3 indicates that of the three possible reassignments that gives an improved changeability, the same μ-robustness, and a lower s-robustness relative to CMy, the highest improvement on changeability—1—is found in known deviant codes (1.3%, 0.0%, −1.5%). Dashed edges indicate possible but unlikely transitions, for example, MTu and MVe derived from MNe. Ser was probably introduced before Met in MNe, because reassignments of Gly (code MTu), Thr (code MYe), and stop (code MVe) improve mainly the changeability, contrary to the reassignment of Ser that decreases the changeability. Additional constraints besides the improved changeability on the reassignments Leu → Thr and Arg → stop from Hyp code 1 are respectively increased number of dispatchers with neutral polarity, and improved individual changeability of stop codons. The reassignment enclosed by a box generates a genetic code with 22 phenotypes, found in some organisms, where Sec is the additional amino acid.