Hidden structures
| Case number: | 845799-694110 |
| Topic: | Biochem |
| Opened by: | DisposableHeart |
| Status: | Closed |
| Type: | Question |
| Opened on: | Friday, December 12, 2008 - 11:24 |
| Last modified: | Thursday, February 19, 2009 - 06:10 |
We receive puzzles with certain preset structures. I am not aware of why these structures are there, meaning I do not know what amino acids make up a certain protein.
Many times during folding, however, I can achieve better result if I turn loops into sheets, permanently, even though they were not present in the original structure. I make hydrogen bonds that persist fairly well and contribute to the score.
The questions are:
- Are we expected to figure out these type of tricks?
- Is the structure of molecules completely determined by the amino acids, or depending on the actual deformation loops may really turn into sheets and form Hydrogen bonds?
In general, it would be very beneficial to know the amino acid sequence, because one can learn how the amino acids behave and, for example, do not play with a rigid one for hours to no avail.
Can we have a mode that colors the amino acids and tags them?
| Topic: » Biochem |
| Status: Array » Open |
| Type: Array » Question |
Set details.
Assigned to a biochemist.
Hello DisposableHeart,
You bring up a very good point about the secondary structure assignment used by Rosetta.
The helices, loops and sheets that your puzzles start out with are what Rosetta believes to be the correct secondary structure, but as you already know Rosetta isn't always right about everything!
Sometimes rosetta has a loop that runs just next to a sheet, but it just isn't able to correctly form hydrogen bonds and actually make a sheet.
It is for this very reason that we implemented 'structure mode' to let you change the secondary structure if that improves your score.
I wouldn't say that we are expecting you to figure out these tricks, rather it seems as if you and other foldit players have discovered a trick that may improve predictions!
If you are able to significantly improve your score with this method, then it might be an avenue worth exploring.
If we notice that the resulting models are much further from the native than those models that do not change the secondary structure, we will let you know and change the Foldit scoring function accordingly.
For your last question, you can see each amino acid in 'note mode'.
Thank you, very insightful, indeed.
I think it'd worth a blog post.
Thanks,
Peter
| Status: Open » Closed |
Splendid -- marking as resolved.
| Status: Closed » Open |
I was wondering this. If you looked at a listing of a protein's amino acids should you be able to know something about it's structure? That's a great question. Thanks.
I'm glad this was helpful!
csohad asked David Baker a very similar question on the blog:
http://game.bakerlab.org/portal/node/901725
I thought I could show you an example of a secondary structure prediction:
http://compbio.soe.ucsc.edu/~karplus/casp8/T0499/T0499.dssp-ehl2-logo.pdf
This shows a prediction based only on the amino acid sequence of this particular CASP8 target and it's shown in a sequence logo format.
For each amino acid, there is a predicted letter (H for Helix, E for Sheet, C for Loop) above it. The taller the letter the more highly predicted the secondary structure is for that amino acid.
For this case you can see that the beginning of the protein is highly predicted to form a sheet, then the secondary structure prediction gives each letter an equal probability, then there is a very low predicted helix region, before a highly predicted region... and so on.
| Status: Open » Closed |
Marking as resolved.
I have been wondering the same thing ever since the tutorial showed me the "rebuild" command. I never went to the level of creating sheets, but some rebuilds did seem to make very significant changes. (Since I know no details about the amino acids or proteins).