Is our genetic code the best of all possible codes? A 2000 article concludes that maybe it is, depending on what “best” and “possible” mean in this context. Although the article goes well beyond my expertise, I attempt to explain.
The genetic code in question is DNA’s “language” of 3-base-pair-long codons, each specifying an amino acid for building proteins.
What would the “best” genetic code be? The authors hypothesize that the best code might be the one that produces the fewest faulty proteins if a mutation changes one of the base pairs in the codon. In this code, a change in DNA would either produce specify the same codon or a physiochemically similar one. But how do we measure amino acid physiochemical similarity? There are several plausible possibilities. The authors test a few. The actual genetic code scores extremely well on one, and they go with that, hopefully not circularly.
How does one measure how good a code is at inserting the same or similar amino acids after point mutations? There are several ways to measure that, too. Authors choose one for reasons that are probably plausible, though I wouldn’t know. Along the way they explain that people using other methods found that there could be many better codes only because they used the wrong method.
Now that they have defined “best” and “similar”, what is “possible”? The authors briefly explain what random assignment of amino acids to codons would produce a LOT of errors and choose “possible” to mean either “arangements that remain synonymous codon block structure [of real DNA] for all variants” or organized according to a plausible hypothesis of evolution of the code. Thus they have two different sets data to analyze.
After solving a last problem – how to measure optimality – the authors are ready to report their results.
Results and conclusions: The code is adaptive and apparently the result of selection for minimizing protein errors after mutation errors. Its apparent excellence is not a result of the analytical methods used. Its excellence is not a result of optimizing treatment of a few extreme amino acids, but of how it treats the entire set. The code is not a result of stereochemic limitations. If the possibilities are restricted to those considered evolutionarily likely and if a realistic transition bias is included, the code is 95-100% optimized. It is possible that no better code could develop under these restrictions.
Under any assumptions, the actual code is extremely good and equally good or better codes are extremely rare. Nonetheless, there theoretically can be a few better codes. Most of these probably can’t evolve given plausible evolutionary starting points for the code. Only under evolutionarily plausible restrictions may the actual code be the best of all possible codes.
Source: Freeland, Stephen J., Robin D. Knight, Laura F. Landweber, and Laurence D. Hurst. 2000. Early Fixation of an Optimal Genetic Code. Molecular Biology and Evolution 17(4): 551-518. DOI: 10.1093/oxfordjournals.molbev.a026331