Mutate Puzzle Default Residues

Case number:699969-997092
Opened by:wisky
Opened on:Saturday, February 22, 2014 - 05:00
Last modified:Saturday, March 1, 2014 - 02:37

Why have you been giving design puzzles with all isoleucines pressed into the backbone? It has been brought to my attention that many players might use the default residues on them, and that helices do not rebuild at all with the default residues. Many players might not shift the amino acids from the defaults at the start of the puzzle (or even shake!), think the helix rebuild is wonky, get irritated and not play these puzzles... Alanines, serines, or threonines might be a better choice for default residues.

(Sat, 02/22/2014 - 05:00  |  14 comments)

Susume's picture
User offline. Last seen 15 hours 20 min ago. Offline
Joined: 10/02/2011

I tested this idea that some other amino acid would rebuild helices better in design puzzles. This is a change that could be made on puzzles (by changing the default aa), rather than in the code, and would certainly save people from giving up in disgust. Many people don't look at sidechains at all when they are first shaping a puzzle, and would not think of changing them by hand prior to trying to make a helix.

On 65 residue dimer design puzzle, I started from reset, cut off the first 14 residues, changed to helix, and attempted to rebuild them. I arbitrarily chose 40 as the maximum number of rebuild poses that someone would wait through before giving up on the tool, though for many people it would be fewer. Ran 10 trials per condition, each starting from extended chain.

All isoleucine (as given at puzzle start), no shaking:
7 out of 10 trials failed (did not produce a decent helix in 40 poses or less)
number of poses required on successful trials: 13, 14, 19

All isoleucines, shake 2 iters before starting rebuild:
6 out of 10 failed
successes: 12, 12, 13, 35

All alanines:
4 out of 10 failed
successes: 7, 7, 17, 19, 23, 39

All glutamates:
1 out of 10 failed (succeeded at 44 poses)
successes: 5, 5, 6, 6, 6, 7, 8, 14, 26
note: with glutamate, whenever it took more than 10 poses to get a good helix, it was only the last 2 segments that would not curl up

Obviously the devs could test other amino acids (methionine seems an obvious candidate, if hydrophobes are preferred). The downside of picking one that has a high affinity for helices might be that it makes rebuilding sheets harder, but currently people in chat are complaining about helices, not sheets.

spmm's picture
User offline. Last seen 11 hours 27 min ago. Offline
Joined: 08/05/2010
Groups: Void Crushers

Rebuilding anything particularly helixes in design puzzles was difficult for some time before NC in my experience. Particularly longer helixes. Usually would form shorter helixes wrapped up on themselves which was sometimes worthwhile. Always needed extra work. Isoleucines don't seem to make much difference except in the loops imho. Seem to provide a better distance as well when moving the sym parts together, but then I am not very good at design puzzles.

Joined: 04/11/2013

I used Tweak, Idealise and rebuild and ended up just tweaking. Hope this helps.

bkoep's picture
User offline. Last seen 15 hours 41 min ago. Offline
Joined: 11/15/2012
Groups: Foldit Staff

You bring up a good point—that poly-Ile is not very good for early design techniques (like rebuilding a helix)—but I'm not sure this is something we want to change.

Our hope is that this starting configuration will encourage different approaches to design. If we optimize the starting sequence to favor helices, then I'm afraid this would provide a clear "path of least resistance" to helical solutions. A lot of people (especially new players, who are still learning the Foldit tools) could inadvertently fall into a pattern of making helices for every design puzzle. And while helical designs are fine, and often score very well, we're much more interested in the breadth of designs that different players can invent. To this end, we want to avoid catering to any particular style of design.

Advanced players are free to mutate to Glu and rebuild helices, and to share techniques with players who are frustrated by poly-Ile rebuild results. But we'd also like to encourage newcomers to develop their own strategies for molding a poly-Ile chain into a high-scoring solution.

Joined: 05/18/2012
Groups: Contenders

I would have to disagree bkoep. By your own logic, you ought to provide a starting arrangement that makes it easy to work with both helices and sheets. By using poly-lle as the default, in addition to NC problems building helices, you are predjudicing people against using helices in their designs altogether.

While there may be better choices, I've found that working with alanines at early stages makes it easy to build both helices and sheets. There are a few scripts that allow us to input the primary sequence, and while I appreciate the efforts of the programmers that have written them, it still aggravating to have to use these scripts at the beginning of every monomer or symmetrical puzzle to change the default arrangement.

spmm's picture
User offline. Last seen 11 hours 27 min ago. Offline
Joined: 08/05/2010
Groups: Void Crushers

If you use the selection interface you can simply select all, click the M button and choose which AA to change them all to. Only if you want to do a global replace of course.

Joined: 05/18/2012
Groups: Contenders

Thanks spmm. That may be helpful for others to know. Unfortunately, I find the select interface non-intuitive and complex. Somewhat akin to using vi instead of a full screen editor.

Joined: 04/19/2009

This is a perfect example of the difference between the dev point of view - "hoping that this starting configuration will encourage different approaches to design" - and the players point of view, that "Many players might not shift the amino acids from the defaults at the start of the puzzle (or even shake!), think the helix rebuild is wonky, get irritated and not play these puzzles".

This is why we need to have a very long and leisurely dev chat about this client. There has to be a happy medium that only discussion with the whole (dev and player) community can resolve, on many of the client issues.

Joined: 09/19/2011
Groups: None

What you're asking for doesn't exist.

Yes. It's true. Players tend to not mutate the original sequence. If the devs insert any bias, the players will just fold around the provided start instead of designing.

The default was switched to isoleucine because alanine does not form a stable core. Rosetta wouldn't mutate to large sidechains and the players didn't either.

Players need to be designing both the PS and SS but they don't have the knowledge, or the time, or the tools. People are complaining now, because the PS is interfering with rebuild and thus wasting their time, but they weren't changing the PS enough before, either. (I admit, I don't design the PS either, beyond running mutate combo).

This has been going on for years. There must be over 100 postings about rebuild and helices. There are two tools that need to be added to make up for the fact that is supposed to be a game played by non-scientists and not R@H brute-force-rebuilding. I'll add them as separate feedbacks today.

Joined: 05/18/2012
Groups: Contenders

Thanks brow42. I thought I'd elaborate here on my choice of using alanines. While it may not form a stable core (I'll take your word for it, since I have no idea), I only use them until I have the monomer design I'm testing. Once it is set up, then I run mutate and/or mutate scripts to determine if the result has a stable core. My choice of alanines is to simplify the cut/paste/SS formation stages only. Once a design with a stable core is found, I worry about joining the n-copies and further mutates.
I did in fact, worry whether the starting choice of alanines had any effect on my "final" protein, but it looks as if it resulted in only 2 alanines. I would hope that 2/65 is not undesirable, but again, I have no idea.
This is all based on my observations, but if anyone can explain that this is undesirable, and why, I would appreciate it. :)

wisky's picture
User offline. Last seen 11 weeks 1 day ago. Offline
Joined: 07/13/2011

From my experience all alanine starting poses tend to lead to many glycines inside the core as well as alanines...

Joined: 09/21/2011

I'd like to convert a string of segments to helix structure, then double click on that stucture and have it instantly coil. i'd like to double click a sheet and have it instantaneously straighten. don't care what the aas look like cuz i usually gotta tweak it till the cows come home regardless.

Joined: 09/21/2011
Groups: Void Crushers

This would interfere with the selection mechanisms in selection mode, idea nice implementation not this way.

LociOiling's picture
User offline. Last seen 10 hours 15 min ago. Offline
Joined: 12/27/2012
Groups: Beta Folders

See also, suggestion that design goals be enforced via filters instead of by messing with the rebuild function.


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, RosettaCommons