Rotation affects score when it shouldn't. When filters are present (Puzzle 691 residue tetramer)

Case number:845813-994915
Topic:Game: Tools
Opened by:wozzarelli
Status:Open
Type:Bug
Opened on:Saturday, April 6, 2013 - 16:34
Last modified:Friday, April 12, 2013 - 16:01

Showed on Mac version (up to date as of 6/4/13).

Occurred in Puzzle 691 - 60 Residue Tetramer

Ensuring each protein in the tetramer was separated, I was able to translate the proteins around within the screen without affecting the score. This was expected. See uploaded files:
Screen Shot 2013-04-07 at 2.28.44 AM.png
Screen Shot 2013-04-07 at 2.37.39 AM.png

I then rotated the protein in place, which significantly affected the score. This was unexpected. See:
Screen Shot 2013-04-07 at 2.50.23 AM.png

Subsequent translations of the protein within the plane did not affect the score further. But another rotation then affected the score still further. See
Screen Shot 2013-04-07 at 3.04.42 AM

Notes:
1. The values for the last three filters: Core Existence, Polar Core and Layer Design were all seen to change. In all cases the residue IE score was met so it is unknown if this filter would have been affected.
2. I was careful to ensure that the protein shape was unchanged on rotation and was not accidentally pulled. To verify this, note that the score in each case is 14076.373 which seems to indicate the filters are the source of the problem.
3. Did not occur in:
Puzzle 693: Hand-Folding De-novo Freestyle 22, for which there are no filters.
Puzzle 694: CASP T0711 Disulfide Repost Round 2 which has filters but no penalties. My protein had no disulfide bonds at the time.

AttachmentSize
Screen Shot 2013-04-07 at 2.28.44 AM.png145.89 KB
Screen Shot 2013-04-07 at 2.37.39 AM.png142.12 KB
Screen Shot 2013-04-07 at 2.50.23 AM.png153.43 KB
Screen Shot 2013-04-07 at 3.04.42 AM.png159.36 KB
(Sat, 04/06/2013 - 16:34  |  9 comments)


Joined: 09/21/2011
Groups: Void Crushers

Sorry, I looked at them all, all scores are the same.

Joined: 09/21/2011
Groups: Void Crushers

And priority down to 3, this is not a game breaker.

Joined: 05/27/2010
Groups: None

Well that was embarrassing... I must have attached the incorrect images.

Please find attached two images, with the only difference being that the proteins have been rotated in place.

Joined: 09/21/2011
Groups: Void Crushers

So it looks like the score function on sym really likes the parallel aligning, even if the distance is big.
Do you also see the effect when the distance between is really huge?

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

assigning

Joined: 05/27/2010
Groups: None

Yes, I'm still seeing that effect over long distances.

Is there is an element to the scoring that looks for parallel alignment? There might be situations where a single protein has helices in several directions?

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

Both 'core existence' and 'polar core' look at the surface area of the protein. They take into account the symmetric units in the surface area calculation, so when you rotate your protein the residues it considers in the 'core' are probably changing. This will change your score for both 'core existence' and 'polar core'.

If this is what's happening then it's not a bug. If you want us to look further at the issue, then could you save your solution and share it with yourself. Then just post the name of the solution here and we can look at it.

Joined: 05/27/2010
Groups: None

My observation was that once the four proteins are sufficiently separated, the residues that form part of the core should be rotation independent.

I've saved a local solution, "Separate and Rotate" and tried to "Upload For Myself", but there's an error getting shared solutions from the server. So I'm not sure if the upload succeeded or not.

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

Thanks, I'll look into it

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