Symmetry layer test; Mutable detection fails on test puzzle

Case number:845813-993855
Topic:Game: Tools
Opened by:Bletchley Park
Status:Open
Type:Question
Opened on:Tuesday, November 6, 2012 - 19:45
Last modified:Tuesday, November 13, 2012 - 21:05

The mutable detection routines no longer work properly on this latest-client requiring test puzzle.

Number of mutables detected is 1.

(Tue, 11/06/2012 - 19:45  |  23 comments)


Joined: 05/19/2009
Groups: Contenders
Type: Bug » Question

ok, changing the test sidechains will make it work again, but there should be a simpler way to test for mutability.

tamirh's picture
User offline. Last seen 6 years 19 weeks ago. Offline
Joined: 05/11/2012

assigning

brow42's picture
User offline. Last seen 6 days 11 hours ago. Offline
Joined: 09/19/2011
Groups: None

I haven't tried this (is it a devprev only puzzle?) but in principle the IsMutable() Lua2 function should be doing it, instead of mass mutate-range (employed in Lua1).

Which algorithm/script are you using?

Joined: 05/19/2009
Groups: Contenders

Hi Brow, the set-get set-get compare method. Where is a pointer to all LUA1 and 2 functions ?

Joined: 05/19/2009
Groups: Contenders

And yes it is a devprev puzzle for which you need the latest devprev client.

spmm's picture
User offline. Last seen 41 weeks 3 days ago. Offline
Joined: 08/05/2010
Groups: Void Crushers

it is a bit odd because the scripts can only see 1 or two mutables when in fact they are all mutable.

brow42's picture
User offline. Last seen 6 days 11 hours ago. Offline
Joined: 09/19/2011
Groups: None

A truly bizarre bug and a lucky catch, thanks BP.

SetAminoAcid and SetAminoAcidSelected do not mutate to all AAs. It fails to mutate the selection to: a,c,f,i,m,l,w,v,y. It can mutate to e,d,g,h,k,n,q,p,r,s,t. SetAminoAcid can mutate the first segment. It seems GUI mutate is similarly restricted. Once I make a hydrophobic core, it is possible to mutate those segments to anything.

The Lua 2 functions are at http://foldit.wikia.com/wiki/Foldit_Lua_Functions and http://foldit.wikia.com/wiki/Lua_Standard_Libraries.

tamirh's picture
User offline. Last seen 6 years 19 weeks ago. Offline
Joined: 05/11/2012

If you are seeing the restrictions in both the GUI and the script this is the intended behavior. This is one of the tools we're testing out to prevent things like having too many Tryptophans on the outside of proteins (among other issues that come up with designs).

If you select more than one protein to mutate it will only give you the most restrictive selection. For example, if AA #1 can not mutate to X, Y, or Z and AA #2 can not mutate to A, B, or C and you select #1 & #2, you will not be able to mutate to A, B, C, X, Y, or Z.

If your scripts select a large range of amino acids and try to see if they're mutatable it's very likely that the individual restrictions combined makes it so that there are no valid choices for the entire selection.

Joined: 09/21/2011
Groups: Void Crushers

Why not implement the restrictions on a segment base instead of the sum of all segments? This will break a lot of mutating scripts.

tamirh's picture
User offline. Last seen 6 years 19 weeks ago. Offline
Joined: 05/11/2012

We thought it was more clear to have the list of allowable types be the same as what will actually work with your selection. This gets around the outcome of having only 1/2 or some fraction of the mutations actually work. I agree it doesn't give you a good indication of what is doing the restriction so you can narrow down your selection.

It would be possible to allow the mutation where it's not restricted and pop up a message telling you X/Y were actually mutated. Anyone else have an opinion on which way they like?

jflat06 also just pointed out to me that there's no LUA function to tell if a certain residue can mutate into a certain AA. I'm adding that in now.

brow42's picture
User offline. Last seen 6 days 11 hours ago. Offline
Joined: 09/19/2011
Groups: None

Mutation selection wheel does not fade completely until the mouse is clicked (original interface).

tamirh's picture
User offline. Last seen 6 years 19 weeks ago. Offline
Joined: 05/11/2012

That's normal. It was like that before as far as I can remember yes?

Joined: 09/21/2011
Groups: Void Crushers

If you design in parts and join later, outside can become inside. Especially in symmetry designs restristions based on inside-outside would restrict the designs very much. Giving penalty points instead of restricting the possibilities is imho a better way to go.

tamirh's picture
User offline. Last seen 6 years 19 weeks ago. Offline
Joined: 05/11/2012

We have a filter that adds/subtracts points based on this same calculation. We're just testing them separately.

Susume's picture
User offline. Last seen 1 week 6 days ago. Offline
Joined: 10/02/2011

As someone who often works with the protein pulled apart, I second this concern that what is temporarily outside may be intended to become inside. Just as you gave us a checkbox to disable cutpoint constraints, maybe you can let us disable the AA filter.

tamirh's picture
User offline. Last seen 6 years 19 weeks ago. Offline
Joined: 05/11/2012

When you are working with the protein as a chain, are you doing mutations at that point? I was thinking it wouldn't matter because people would modify the backbone first from a chain and then they would be doing mutations. If I'm wrong there, then something will probably have to change. We may have to modify the filter somewhat to temporarily be turned off if the protein is pretty much still a chain.

Adding a checkbox for this filter would defeat the purpose of it, because you can strictly get more points with it off. What it does is try to prevent designs that are not possible to make. We have a score filter that takes this fact and modifies the score, but we wanted to have both tools at our disposal.

We thought of having this only as a final 'verify solution' button that you would have to press to have a valid score, but that didn't seem like a very good idea either. If you got a very good score with the checkbox off and then turned it on at the end, you may find your protein scores poorly once you've fixed it to have the correct AAs based on the filter.

Joined: 05/19/2009
Groups: Contenders

I have hardly ever taken a backbone apart in mutational puzzles. I always put a backbone in place, then start mutating and modify the backbone again, it is a repetitive exercise. I rely on the software to choose the correct / most suitable sidechains based on the current backbone configuration.

I also notice that when there are severe penalties in score based on backbone / sidechain combinations, mutate will not start mutating the lowest scoring segments, you might consider giving segments priority low scoring to high scoring ?

brow42's picture
User offline. Last seen 6 days 11 hours ago. Offline
Joined: 09/19/2011
Groups: None

The filter seems harsh. The straight segment disallowed A but allowed G. Later I found even G disallowed. Surely those two should always be allowed.

tamirh's picture
User offline. Last seen 6 years 19 weeks ago. Offline
Joined: 05/11/2012

This test puzzle is just using the default values for the filter, it's more to test the functionality and not the specifics of it.

For scoring puzzles the Bakerlab scientists will customize the filter to work more specifically on exactly what they want to encourage/prevent.

As for having a specific AA disallowed at one orientation and not at another. That's just going to happen as a result of having to have some cutoff point.

Susume's picture
User offline. Last seen 1 week 6 days ago. Offline
Joined: 10/02/2011

I cut into pieces, assign sheets and/or helices, assign aas in certain places, rebuild as needed, then reassemble. If I am planning hairpin turns I assign appropriate aas before reassembling; if I'm planning a hydrophobic core I assign small hydrophobes before reassembling, and if I am planning hydrophobic interactions between monomers I assign small hydrophobes after assembling but before pulling the monomers back together. I don't use the mutate function until much later, but yes I do use design mode from the very beginning. I may just have to learn to do things in a different order.

(sry, did not mean to down-vote your comment, just clicked wrong)

Susume's picture
User offline. Last seen 1 week 6 days ago. Offline
Joined: 10/02/2011

All the disallowed aas (orange and blue) are giving the same message when I attempt to change to them: "Cannot place a hydrophillic amino acid with this shape at this residue" - I assume there is supposed to be a different message for the hydrophobic ones. (Also, hydrophilic has only one L.)

Disallowing aas based on current protein shape prevents the tactic of setting an aa to something new and then rebuilding, wiggling etc. to make it fit. I have yet to profit from this tactic, but I do sometimes try it. I hate to see that kind of exploratory play prevented.

Different issue: I have an orange valine pointed inward, and all blue aas are now disallowed for that position. I'm not sure this is correct, as a blue might be good in that position *if* it forms appropriate hydrogen bond(s). Does the filter allow for that?

tamirh's picture
User offline. Last seen 6 years 19 weeks ago. Offline
Joined: 05/11/2012

Looks like it's not showing the correct message in the original interface. I'll fix that and the typo.

I think we might end up with using the score version of this filter (which doesn't explicitly prevent mutating into different aa's it will just score them poorly).

As far as your last question, I'll forward it on to the biochemists, I'm not sure.

spmm's picture
User offline. Last seen 41 weeks 3 days ago. Offline
Joined: 08/05/2010
Groups: Void Crushers

I have been running a 'boring helix' on 40 Res Retry and am, not unusually, confused.

In screenshot one I have 3 hydrophobic residues which look 'proper', ie rings parrallel, but have little red glows and an IE penalty:

http://fold.it/portal/files/chatimg/irc_179219_1352839074.png

In screenshot 2 I have switched the 3 glowing hydros to lysine, along with the expected drop in score there is a larger glow but no IE penalty:

http://fold.it/portal/files/chatimg/irc_179219_1352839400.png

You can hopefully see from the view menu that voids are turned off and exposeds are on.

I was expecting to have a 'blue orange orange' or 'orange blue blue' repeat pattern emerging like a classic helix but it seems to prefer all hydrophillics.

I am not really bothered about the points but I would like to understand how to get a classic, basic, boring helix right in a couple of configurations as there are so many of them in the puzzles :)

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