The modern genetic code of 64 3-letter codon ‘words’ for 20 amino acids and 3 Stop signals is universal across all the organisms in the Tree of Life.  Its evolutionary origins are still mysterious, however.  It is obvious that the code could not have sprung into existence ‘as is’; meaning that its complexity must have evolved through several simpler ancestral codes.  Recent work using ancient and likely highly conserved protein sequences provides a potential order in which amino acids were added to the code, but this work by itself does not explain which codon ought to be paired to which amino acid.  Nor, does this order explain when and where the Stop codons were added.  Such Stop codons are essential for ending protein construction, but they are also likely to be the most catastrophic for protein function when mutated.  Adding a Stop signal in the middle of a coding sequence ‘breaks’ the produced protein.  Removing the Stop adds a superfluous string of amino acids. Therefore, if a Stop codon was an early addition to the evolving genetic code, subsequent additions of amino acids could have preferentially paired with codons in a pattern that minimizes the frequency of mutations to or from Stop.  Nonacs and Nonacs (2025) modeled the sequential evolution of the genetic code if an early Stop signal was also present: I.e., the Catastrophic Mutation Minimization Hypothesis (CMMH).  Multiple predictions from the CMMH are consistent with the order in which codons pair with amino acids.  Most strikingly, the amino acid cysteine is accurately predicted to occupy the closest mutation location to Stop codons.  Pairs of cysteines can form disulfide bridges that are essential for protein function, and therefore, can also result in catastrophic mutations.  Thus, the CMMH predicts that a Stop/cysteine codon ‘neighborhood’ is both essential for protein construction, and simultaneously mutationally most removed from the rest of the code.  In total, 16 of the 20 amino acid to codon pairings are predicted by the CMMH – a precision heretofore not achieved by any other model of genetic code evolution.  

https://doi.org/10.1007/s00239-025-10294-0