Mutable Locked Segments on 537

Case number:845813-992321
Topic:Game: Tools
Opened by:brow42
Status:Open
Type:Bug
Opened on:Sunday, April 8, 2012 - 23:45
Last modified:Tuesday, April 10, 2012 - 03:07

I noticed that many locked segments on 537 CO2 and You Too were mutable via Lua 2, but not via the interface. This is unfortunate, since many of the locked protein segments are low scoring. For example, 298 Histadine scores -63...after mutating to Glycine via Lua it scores +9...that one change took me from rank 50 to rank 24. Non-scriptors, or even infrequent scriptors, may never have known they were mutable. Plus, given the size of the puzzle, many scripts might not even get to that segment, and there's no manual recourse to fix that horrible segment (I found it only by writing my own script).

Is this intentional i.e. you wanted the backbone locked but the sidechains mutable, and this is an unfortunate side effect?

main client 2012-02-24-47632-win_x86

(Sun, 04/08/2012 - 23:45  |  16 comments)


Joined: 06/17/2010

pdb #298 his - on dev it is locked and not mutable.
same on main client when loaded my best solution and after reset - cant mutate, it is locked.
Can`t reproduce it, share your solution brow!

--- EDIT
Just got #1 in tis puzzle when started http://fold.it/portal/recipe/40448 brow`s recipe on my best solution.
It is somehow moving locked sidechains...

Joined: 06/17/2010

Log form Brow`s script started on my #7 solution.
It is mutating locked parts.

brow42's picture
User offline. Last seen 5 days 17 hours ago. Offline
Joined: 09/19/2011
Groups: None
Type: Question » Bug

We've determined that Rav's scripts do not see this, because he only mutates selected AAs, and locked segments cannot be mutated. I mutate by segment number. I find 215 mutable, Rav finds fewer.

count = 0
for i = 1, structure.GetCount() do
if structure.GetSecondaryStructure ~= 'M' then
if structure.IsMutable(i) then print(i) count = count+1 end
end
end
print('Total mutable:',count)

Joined: 06/17/2010

It is combination of 2 bugs.
This puzzle returns as mutable 215 segments where 120 of this segments are locked.
My mutate script is using "select" then "structure.SetAminoAcidSelected(type)".
Brow is using "structure.SetAminoAcid(number,type)".
Because locked can not be selected my script is not touching them, Brow`s script is directly using mutate on segment and puzzle setup ALLOWS mutation of locked segment.
Attached log and script.

edit:
It is working on DEV, MAIN still not have IsLocked functionality. Main also returns 215 mutables.

Joined: 06/17/2010

Loaded 529 - same thing can be done in this puzzle.
Tlaloc`s and Brow`s and any v2 script that are using mutate(sn,typ) broke it.

Joined: 06/17/2010

http://fold.it/portal/recipe/40457
Script that check solution have mutated locked segments or not.

brow42's picture
User offline. Last seen 5 days 17 hours ago. Offline
Joined: 09/19/2011
Groups: None

215 IsMutable = true
120 IsMutable and IsLocked = true
101 true mutables found the old way! O_O

If IsLocked is to be believed, Mutate Combo mutated up to 6 locked segments (537 clean script says my rank 100 score is clean, but it has 2 mutated locked segments).

Let's see, the extra segments marked true mutable are 38, 220, 225, 301, 316, and 319...VLYHLL. A script shows these 6 as both selected and locked!

list= { 38, 220, 225, 301, 316, 319}
selection.SelectAll()
--structure.SetAminoAcidSelected('g')
for i = 1,#list do
print(list[i], selection.IsSelected(list[i]), structure.IsLocked(list[i]), structure.GetAminoAcid(list[i]))
end

brow42's picture
User offline. Last seen 5 days 17 hours ago. Offline
Joined: 09/19/2011
Groups: None

Tracking all these down shows that they have locked backbones and free sidechains (whew).

Request structure.IsLocked() return a 2-tuple for BB and SC status, or a 3-tuple = BB and SC, BB, SC.

beta_helix's picture
User offline. Last seen 1 day 4 hours ago. Offline
Joined: 05/09/2008
Groups: None
Status: Open » Open

Thanks for catching this brow42 & Rav3n_pl.
We believe that it is only an issue with the CO2 puzzles, but we will try to figure out why only the Lua V2 scripts were able to find this loophole!

I also replied in the other thread:
http://fold.it/portal/node/992322#comment-16706

brow42's picture
User offline. Last seen 5 days 17 hours ago. Offline
Joined: 09/19/2011
Groups: None

I wrote a diagnostic program to help catch this. It's updatable for future puzzles. It requires devprev and IsLocked, thanks for adding that function.

Good News: I used IsLocked so my script didn't make bad mutations for people using devprev.

Bad News: I have a rank 100 solution from the day before (which I uploaded to server), that only used GUI Mutate Sidechains, Raven Mutate Combo 1.4, Tlaloc Mutate All 4.0, and it has 2 bad mutations.

http://fold.it/portal/recipe/40456

537: CO2andYouToo!	04/08/12 23:29:00
Locked:	     (1 - 28), (37 - 52), (70 - 114), (125 - 164), (174 - 242), (255 - 368), (393 - 402)
Not Locked:  (29 - 36), (53 - 69), (115 - 124), (165 - 173), (243 - 254), (369 - 392), (403 - 418)
Mutable:     (3 - 21), (29 - 45), (52 - 55), (57 - 70), (114 - 124), (126 - 129), (165 - 178),
             (219 - 233), (241 - 260), (265 - 276), (284 - 306), (315 - 323), (329 - 333), (335 - 338),
             (368 - 393), (401 - 418)
Not Mutable: (1 - 2), (22 - 28), (46 - 51), 56, (71 - 113), 125, (130 - 164), (179 - 218), 
             (234 - 240), (261 - 264), (277 - 283), (307 - 314), (324 - 328), 334, (339 - 367), (394 - 400)
Mutable But Locked:	(1 - 2), (22 - 36), (46 - 51), (53 - 69), (71 - 113), (115 - 125), 
             (130 - 173), (179 - 218), (234 - 240), (243 - 254), (261 - 264), (277 - 283),
             (307 - 314), (324 - 328), 334, (339 - 367), (369 - 392), (394 - 400), (403 - 418)
Mutable and Not Locked:	(29 - 36), (53 - 55), (57 - 69), (115 - 124), (165 - 173), 
             (243 - 254), (369 - 392), (403 - 418)
Found bad mutations:	225, 319
brow42's picture
User offline. Last seen 5 days 17 hours ago. Offline
Joined: 09/19/2011
Groups: None

Stupid Typos. Just to avoid any confusion, Mutable But Locked above is inverted. Correct list:
Mutable But Locked: (3 - 21), (37 - 45), 52, 70, 114, (126 - 129), (174 - 178), (219 - 233), (241 - 242), (255 - 260), (265 - 276), (284 - 306), (315 - 323), (329 - 333), (335 - 338), 368, 393, (401 - 402)

spmm's picture
User offline. Last seen 36 weeks 4 days ago. Offline
Joined: 08/05/2010
Groups: Void Crushers

the puzzle has points and they seem to be on player pages these are quite wrong.

beta_helix's picture
User offline. Last seen 1 day 4 hours ago. Offline
Joined: 05/09/2008
Groups: None
Status: Open » Open

spmm, do you mean that it's wrong that there should be any points awarded because of the bug?
or
that points were awarded normally since the puzzle closed, but something is wrong with how this puzzle was scored?

Since the bug was noticed so late, we didn't close the puzzle or reduce the number of points or anything. Turns out that if we closed it and re-opened it with the bug fix you would still be able to load in incorrect solutions because there is no check for that (so we'd have to update both clients... etc. you can see it gets messy!)
So the puzzle should have closed as normal and points assigned the way the usually do.

Let us know if something went wrong with the normal scoring!

Thank you and again our apologies for this, but thank you all for catching this mistake!

spmm's picture
User offline. Last seen 36 weeks 4 days ago. Offline
Joined: 08/05/2010
Groups: Void Crushers

The puzzle closed and points were accorded normally.

The problem is that several people leapt on and started using the script even though they knew it was a bug, these players gained massive points from the script, Deen said she got 150 points in about ten minutes.
Other people stopped playing at that point with around 9 hours playtime left. The players who were using the script just kept it running, and of course many people would have been running other scripts and been unaware of the problem.

So the scores just before the script was introduced are fair and could be used, but may also inadvertently contain incorrectly mutated segs - and I assume that your comment in global that: ''Seth says we should be able to scan all the solutions for the correct sequence and give credit based on that.'' will somehow eliminate all of the bad mutations, and may of course penalise people who unknowingly mutated a frozen seg, because their pose and sidechain scores may be adversely impacted. That is ok just bad luck.

Rav and brow have created a test script to run to check the puzzles :)

Joined: 04/19/2009

@ spmm...

It was actually much worse, I let the script run it's course as I said in global, because I still think that something is not quite right with my mac devprev client (I wanted to see if it picked up any normally mutable sidechains and what it did with them). The final tally of points for me was more like 350, and 50 ranks.

I did a pm to beta today stating my concerns about this, and that of course my score should not stand.

As you say - many folders had likely inadvertently run one of the older scripts that also showed this bug in main, both on this puzzle and on the previous CO2.

I do hope that the devs are able to scan the solutions for the bad mutations. There are many folders who worked very hard on these puzzles, and should get the credit they deserve for them.

beta_helix's picture
User offline. Last seen 1 day 4 hours ago. Offline
Joined: 05/09/2008
Groups: None

I have brought this up in a news post: http://fold.it/portal/node/992332

We will manually recalculate all the scores for both puzzles, where only valid solutions will count.

jflat06 is looking into resolving this obscure bug so that it never happens again!

Until then, thanks to Rav and brow for creating a test script to double-check new puzzles!
Please PM me (as well as posting a feedback) if you ever notice this to be a problem in the future.

Sitemap

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