Some of mine are:
The question: how do you optimize the ribosome? The problem: how do we use these strategies to answer that question. Thanks for starting this thread JR, it is a conversation we need to have.
I am using all these strategies, focusing mostly on 3. (Clean up red “error” locations if energy diff. decreases.)
I am focusing on strategy #3, clean up red error locations. In particular I try to fix what Omei refers to as global misfolds, which is when part of the sequence incorrectly binds with a sequence from another side of the puzzle, causing large parts of the ribosome to fold incorrectly. I’m not worried about local misfolds, which is when nucleotides (NTs) are bonding in the correct general area but may be off by a couple NTs.
My ideal solution would be eliminating almost all misfolds by changing only one NT.
(1) I start by focusing on target bonds that are not forming (target helix pairs). I sometimes manually “flip” pairs or I will use the https://eternagame.org/web/script/9251108/ Arc Plot Interface (v2.0) booster to help with this, by locating and mutating target pairs that are not forming in the MFE/natural shape. I look for mutations that both reduce the red areas and lower the delta Free Energy. I look for one of the “best” pair mutations that (a) reduces existing mismatches, (b) does not seem to introduce any significant new mismatches, and © reduces the delta Free Energy. I sometimes stop with a “#stem” design that only mutates target helix pairs.
If I cannot progress after a while, I may also progress to (2) looking at target loop bases that are mismatched in the natural form. I then try mutations of these bases to break the mismatch pairs. The same criteria apply: mutations that (a) reduce existing mismatches, (b) do not seem to introduce any significant new mismatches, and © reduce the delta Free Energy.
I will also start from existing designs by other players, try undoing loop base mods, or mods that make minimal/no dFE gains, and then add mods of my own.
Generally at stem ends, I look to replace G’s with A’s by either replacing GC’s with AU’s or the G’s in a GU pair. Usually keeping same purine/pyrimidine orientation. I’ll do this even if delta breaks even, if I can’t find better G reduction delta change options …at least that’s what my latest strategy was! If I see a way to bring closer to target that reduces delta, I’ll often toss that in too. I also generally don’t touch loops that look like they may be active areas.
But how do we make a more stable single purpose ribo?
I like to flip frequently between natural and target mode, weakening stem structures in the former and strengthening them in the latter. The red error locations seem to be a guide only: improvements in the delta value can be made by changing any part of the structure it seems.
another strategy I’ve tried is downloading all current solutions. Then inputting number of mutations and delta for each. Sorting by chg in delta from start/mutation. Selecting top tier solutions with highest chg in delta/mutation and finding combination of changes that create highest chg in delta/mutation. A good job for a bot.
A corollary strategy is after combining, see if you can find low delta/mutation combinations to eliminate - sometimes these become available.
Astromon [4:55 PM]
I think my strategy will be to have designs w/ and w/o red marks, to give voters a variety of typs of designs to test
maybe ill come around after those and make some with a mixture of both.
My strategy has just been to:
Often large energy differences can be made in the purple substructure region, but because they are “neutral” it might be acceptable to lock those off with strong pairs and ignore the change counter and ignore the energy difference increase.
So my new strategy to lower the energy difference is as follows:
Using all the allowed “change” slots, get the energy difference as low as possible manually. e.g. 18/18 changes.
Go over the 18/18 limit by 1. e.g. 19/18 changes, making sure the energy difference is still lower than the 18/18 string.
Load in the script that jandersonlee provided in this post (“Mark Mutations”) that automatically uses the mark tool to mark modifications to the original puzzle structure.
https://getsatisfaction.com/eternagame/topics/current-lab-modified-base-indicator
Then run the already provided in-game Mutation/Submission tool that will run through individual permutations of all marked bases.
Using the z + y (undo + redo) buttons on keyboard, search through the permutations until you find one that has 18/18 change slots filled and is a lower energy than the original 18/18 in step 1.
Submit solution.
This new string is unique from the original 18/18 string in step 1 and the 19/18 string in step 2, so you can run another Mutation/Submission task for free to see if you can find a lower energy 17/18 or 18/18 string. If you can’t just go back and repeat step 1.
So far I have found this to be a good evolution driven approach to lowering the energy difference.
Brourd mentioned that our 23S designs may not have performed as well as the 16S designs because the 23S subunit is the catalytic center. We also know that one of the best 16S designs placed mutations in only one region of the 16S, specifically in one hairpin. Therefore I suggest we test designs that make mutations in only one domain of the 23S.
There are six regions to the 23S and they largely fold independently. We probably should avoid or be extremely careful in domain 5, the home of the peptidyl transferase center (PTC) where the Big Action happens. I’m going to put a hashtag in my design title indicating which domain and helix I am altering, such as #D1 #H18 #H21. Players using a different design approach may still want to designate the domain or helices in the design title for quicker data analysis. (Do we want to use Roman numerals in our hashtag? I’m not a Roman numeral fan.)
Here is Jieux’s successful 16S design that mutated nucleotides exclusively in one hairpin, before and after mutations:
Using Eterna to fix local misfolds may have helped the 16S fold better. I’m now more inclined to focus on the local misfolds and not worry about the global misfolds (which may not be modeling correctly as I wrote on another forum thread).
If anyone wants to try pasting solutions from the Ribosome Pilot Challenge Warm-Up puzzles into the full ribosome (an approach I’m trying with so far not a great deal of success) the following info showing the nucleotide ranges of the domains may be helpful. Apologies if it duplicates info available elsewhere.
16s
Domain 1 (9-556)
Domain 2 (560-912)
Domain 3 (921-1396)
23s
Domain 1 (18-562)
Domain 2 (562-1268)
Domain 3 (1271-1647)
Domain 4 (1648-2009)
Domain 5 (2043-2625)
Note Domains 1 and 2 of the 23s puzzle overlap by one: I’m guessing the ranges were expanded for the previous puzzle set.
This is what I came here looking for, thanks!
I wish Eternagame could show the 23S subunit like the diagram above. Eli’s pilot challenge design 2.17 mutated 239C to a U. That took the area from 212 to 240 and made a hairpin where the diagram above shows that area part of an open loop & partial hairpin. I can’t tell if it made big changes in the rest of the molecule. 
Oh, and the hairpin is just about in the middle of Domain 1.
I have repurposed the mutation function I wrote to find the IUPAC locations for the 23S full subunit. Here is a link to a google Doc that you can copy & paste to have a list of the IUPAC codes and positions in the molecule as you make your designs.
https://docs.google.com/document/d/1IJqV87LuTtN6HfFr1lVnwsxWr7E87TvohOTAUShFMvE/edit?usp=sharing
Hope this can be helpful.
Do you have a list for the 16s?
@stevetclark Just amended the Document to include the 16S subunit. The link should be the same. Thanks for asking. I should have done both. What was I thinking:)
