I’m thinking part of the reason that humans seem to do so much better than the bots is that humans tend not to make ‘junk’ nucleotide data. What I mean is that humans tend to avoid mutating unpaired bases (such as those contained in a 7-1 bulge or a 6-6 loop), where a bot would paint them and create that junk data. People even avoided mutating the unpaired bases in player projects where I asked them to mutate all bases from adenine! So, I’d like to make a strategy based off of this principal.
Reward a design for each consecutive unpaired Adenine base that is not adjacent to a paired base. (Ex: In a 6-6 loop, there would be eight such ‘rewardable’ bases) Alternatively, punish those designs which mutate non pair-adjacent unpaired bases.
I’d also like to try a second version of this principal, which rewards a fold for each non-mutated unpaired base, regardless of its adjacency to a pairing.