The new "Upload for Scientists" button

We are very excited about the new "Upload for Scientists" button that has been added to Foldit.

As mentioned in a recent blogpost we needed a way for you to send us your favorite Foldit solutions, since those might not be your top-scoring ones.

We wanted to let you know that we haven't only been looking at your top-scoring design solutions this whole time.
We've been clustering all your designs to find the top-scoring solutions for the different topologies you have come up with.

We start with the top-scoring Foldit model and then filter all the remaining solutions, only keeping those that are more than a certain cutoff in RMSD from the starting structure. This results in the top-scoring structure being the only representative model for that particular topology. We then take the next-best scoring Foldit model (that hasn't already been filtered out) and do the same thing… and so on.

In the end, what we are left with are different folds (how different depends on the cutoff we use, we usually try a bunch of different values for a specific Foldit puzzle) but each representative fold is the top-scoring Foldit solution for that shape.

That is how we selected which designs from Puzzle 691 to test experimentally in the lab (see David Baker's blogpost about this), in addition to the ones that you sent to us directly. We will keep you posted on those results soon!

We plan to setup a means of rewarding players who send us useful solutions that are either tested in lab, or used as starting models for subsequent design Puzzles (such as Steven Pletsch's solution that was used as the start for Puzzle 708).

For now, please start using the new "Upload for Scientists" button. For each Foldit puzzle you are able to upload your THREE favorite solutions.
We are excited to see what you send us, particularly for the new Flu Puzzle (718) after chatting with Aaron at last week's Scientist Chat to see what it is he is looking for. If you missed the chat, you can read the transcript here.

( Posted by  beta_helix 120 3144  |  Tue, 05/28/2013 - 06:10  |  3 comments )
3
spmm's picture
User offline. Last seen 3 hours 3 min ago. Offline
Joined: 08/05/2010
Groups: Void Crushers
sanitised transcript

betahelix Hi everyone, thank you for coming to our Scientist Chat!
katfish Hi everybody! We're going to start our scientist chat soon
betahelix bk and Aaron are here from the Baker Lab
Aaron Hi everybody
betahelix Aaron is going to talk about the new Flu Puzzle we just posted... take it away Aaron...

Aaron Yes, we are trying a new approach on this puzzle, trying to find improved stabilization.
Lemme start with a little bit about the old puzzles just to get people warmed up.
So in previous flu design puzzles we were essentially trying to use the foldit players to modify some of our existing binders that bind to H1 Influenza, to bind to H3.
Just so people know, the two most common seasonal flu strains are H1 and H3. Our lab has had H1 binders for a couple years but H3 is a harder target for us.
We essentially docked the old H1 binders to H3 and then had you all try to find mutation that would work against the new bound target.
We had one puzzle where we docked a helix in and one where we tried to complete binder. However, when we tried some of these different helix solutions in the lab they didn't bind; we've tried many helix based designs against H3 but none of them so far have worked.
We are starting to think that though we can get a helix to dock and bind to H1, they just don't work as solutions to H3.
We thought we were very close to an H3 binder with the helix based binders, but alas not, so we originally didn't want to stray too far from what we knew worked for H1.

Ok, so that is a bit of the story on trying to find H3 binders, let me tell you about this puzzle.
One of the things we've found out in lab, through testing, is that though the H1 binders bind very strongly they are not particularly thermally stable their melting temperatures are quite low and one of the side effects of low melting temperature is that they could get degraded quickly by proteases in the body.
Floppy proteins are easy targets for getting chopped up
If we are trying to design a therapeutic, but the proteins get chopped up quickly in the blood, it will never work

So the information we know is:
1) The interface residues on the current binders work quite well
2) The scaffold, the protein the binder was originally designed from, is from a snapdragon flower
3) This means it evolved in temperate climates, and would have never needed to be particularly thermally stable
4) There are similar three helix bundles, like the one in flu puzzles, that came from thermophiles and are much more thermally stable, similar to the one in the flu puzzles
5) The thermophilic three helix bundles usually have a much better packed core than the known H1 binders we have.

SO, if we can build a better packed core for the H1 binder, and have the same interface residues, it could lead to a more stable protein and a better therapeutic.
Aaron that are the design consideration for the puzzle that is posted now feel free to ask me any questions.

Aaron @marie_s: this design binds H1, H2, and H5 (not H3) I'll elaborate, there are three main groups of Influenza, and Influenza B, which is so different it is kind of it's own group. This puzzle is directed towards Group I binders, the most prevalent of which is H1, and why I reference that the most.

TimovdL How do you know if high scoring cores are stable in the way you want it?

Aaron Just so you know, the question that TimovlDL we are discussing right now, it's kind of a slippery question because we know Rosetta/FoldIt has some pathologies that crop up and we want the most stable real designs to score the highest. Ultimately, we won't know the answer until we actually make the proteins and test them for stability

bk @Timo, we also want to make it easier for players to pick their own solutions to share with scientists
bk that's in development right now, and hopefully will be available to you guys soon. We will definitely be looking more closely at the core residues and are working on developing our own filters to score those

Aaron @frood66: We wanted a little more diversity than just locking the backbone and rebuilding the core.

Aaron We decided that allowing a small amount of backbone movement would be suitable but only if we presented the puzzle in the bound state, so the context of the interface is still there
marie_s the 2 little ends are linked to other parts?
Aaron @marie_s: yes actually, they are further aa after those ends, protein tags that lack secondary structure, that we use for some of our experiments

Aaron @janiexq: Short answer no. But well packed cores, with as little free space as possible, correlate strongly with higher melting temperature.

marie_s it is tempting to fold them inside
Aaron those ends needs to be EXPOSED we are hoping the RMSD filter prevents the ends from being folded inside the protein.

So for the flu puzzles, we test them by making yeast cells display the proteins on their surface
So basically both ends do things that you don't see. However, they can be moved a little, but folding them into the center of the protein not the goal. They are like tethers, it's ok, that they are a bit flexible, as long as the ends are free; bk/betahelix/me are all debating right now what do about the ends.

Aaron Our consensus has been not to change the ends, but adding a note on the instruction seems appropriate
Aaron Ok, ya'll I'm going to get back to the bench and making flu proteins

Aaron Thanks so much for you feedback
Aaron The most common flu strains are H1, H3, and Influenza B in that order
Aaron we are moving down the list
Aaron ha, that's what I do
Aaron bye everyone
Aaron and THANKS SO MUCH FOR ALL YOUR WORK!

Joined: 02/08/2012
abuse

isn't there a chance people will abuse this new feature?

Joined: 05/09/2008
Groups: None
How do you mean?

In what way do you foresee someone abusing this, drumpeter18yrs9yrs?

Each player can only upload 3 solutions per puzzle, and it doesn't change your score or ranking in any way.

It's just a way to "Flag a solution for the Foldit team", as the initial feedback suggested:
http://fold.it/portal/node/994901

Get Started: Download
  Windows    OSX    Linux  
Windows
(XP/Vista/7)
OSX
(Intel 10.5 or later)
Linux
(64-bit)

Are you new to Foldit? Click here.

Are you an educator? Click here.
Search
Only search fold.it
Recommend Foldit
User login
Soloists
Evolvers
Groups
Topics
Top New Users
Sitemap

Supported by: UW Center for Game Science, UW Department of Computer Science and Engineering, UW Baker Lab, DARPA, NSF, HHMI, Microsoft, and Adobe