Let us mutate just a few sidechains on denovos like 1319

Case number:699969-2003237
Opened by:Susume
Opened on:Tuesday, December 20, 2016 - 17:03
Last modified:Tuesday, December 20, 2016 - 18:04

Lately we have been getting denovo puzzles that are from foldit designs, which means they may contain quirks that foldit put in. On 1319, I have a hidden blue sidechain (threonine) with nothing to bond to, and I commented to my teammate tokens that foldit should not have put a threonine inside like that, so I must not have folded it right. He pointed out that foldit commonly puts threonine on the inside when you wiggle at low CI and mutate at high CI. My solution would be much more likely to fold up in real life if I could just change that threonine to valine, and the same might possibly be true of the original design.

I'd like to suggest that you give us the ability to mutate denovo puzzles that come from foldit designs, but add a filter that penalizes us if we mutate more than some small number (say 5 or 6) of sidechains. Then we are still working on the designed protein (not designing a new one), but we have the opportunity to improve its sequence just a little, maybe even removing the quirk that keeps Rosetta from being able to predict the designed fold.

If you don't want us to choose where to use our few mutates, a foldit scientist could eyeball the original design and decide if there are any sidechains that look out of place and just make those sidechains mutable in the denovo.

(Tue, 12/20/2016 - 17:03  |  2 comments)

Joined: 09/29/2016
Groups: Gargleblasters

If a noob such as myself may offer another possible suggestion for a way to add mutations...?

I'm sure that with some puzzles there are structural elements which had already been determined and as a result are 'required'. If that is the reason behind these sort of puzzles not allowing for use of Design Mode, I offer a counter argument to that -- which again, I have zero training in this field, so I could be waaay off base with this assumption.

Those predicted ideal sidechains no doubt had been determined either through past lab work or through many computer simulations; however, the flaw which I find in that is when we begin working on them we won't necessarily adhere to the same protein structure, and I believe that is sort of the point behind FoldIt to begin with. To figure out alternative ways for a protein to accomplish the same goal, and possibly be something that neither the past scientists or computer models had came up with. As such, that will also mean that those predicted 'ideal' sidechains will no longer be ideal on account of the various design changes.

As an example: From what I've been able to tell, pretty much all of the non-beginner puzzles have given us access to Structure Mode. If the puzzle begins with there being X-number of Sheets, but we find that converting one to a Helix works better, would it not also be prudent to be able to change the sidechains as well in order to make for the most compact and efficient design possible?

On that note, I offer an alternative suggestion, should Susume's not be viable for one reason or the other...
Instead of having a penalization system, which would add another filter (though, in this case I believe it'd have a negligible performance impact), for these sort of puzzles the Design Mode be 'crippled'. Which just like the penalization system already being in place via filters (through Min/Max number of segments), so too is this from what I've seen in the Tutorial puzzles!
-- Normal/Full 'Design Mode' would operate just as it does now, on puzzles where there are no limits.
-- DeNovo 'Design Mode' would have two main differences. First, since the scoring system already penalizes (in a way) for non-ideal sidechains, we are instead limited to what we can select. In the case of Susume's example, all of the Hydrophobic sidechains would be grayed out, only allowing for selection of different Hydrophilic ones. Second, the right-click "Add/Remove" wheel would either not show up (preferred) or would be completely grayed out.

As I mentioned, those are seemingly already possible from what I can tell. For example, on most segments you may not mutate the sidechain to become GLY, ALA, CYS, or PRO. I assume it's just an IF/THEN check which determines that, and the same would apply here.
IF Puzzle = DeNovo & IF Sidechain = Hydrophilic, THEN Choices = Hydrophilic.
or... IF Puzzle = Design & IF Sidechain = *, THEN Choices = Hydrophilic, Hydrophobic
(Please excuse my rudimentary coding knowledge...)
Similarly, whatever method was used in the Tutorial puzzles to block the usage of the right-click wheel, that would be employed here as well.

Sorry for the long-winded response, I just try to be as detailed and descriptive as I can, to make sure what I'm trying to say is properly conveyed. :P

Joined: 09/29/2016
Groups: Gargleblasters

Dang, screwed up the closing bracket on my Bold :( Should've only read as:
"on most segments you may not mutate the sidechain"


Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Amazon, Microsoft, Adobe, Boehringer Ingelheim, RosettaCommons