Figure 1

Model. Sketch of the model: We start each simulation with the native amino acid sequence as obtained from the Protein Databank (PDB) of the chosen structure, from which we construct a corresponding 'native' DNA sequence by randomly choosing codons using the genetic code under consideration with weights determined by the given AT content (inverse translation). This 'native' DNA sequence is hence as close as possible to the equilibrium with the chosen AT content and becomes the first wild type DNA sequence. Then, at every step, the current wild type DNA sequence is mutated to generate a mutated DNA sequence, which is translated to a mutated amino acid sequence using the genetic code under considerations. The resulting mutated amino acid sequence is evaluated using the folding model with respect to its folding stabilities, based on which the mutated amino acid sequence is either considered as neutral (both unfolding and misfolding stability are above threshold) and the mutated DNA sequence becomes the new wild type DNA sequence, or as lethal (one or both stabilities are below threshold) and the mutated DNA sequence is discarded (the wild type DNA sequence remains as is).