puzzle picture
1108: Compact 37-Residue Marburg Virus Inhibitor Design
Status: Closed


Name: 1108: Compact 37-Residue Marburg Virus Inhibitor Design
Status: Closed
Created: 07/01/2015
Points: 100
Expired: 07/08/2015 - 23:00
Difficulty: Intermediate
Description: Based on the results from Puzzle 1073: Marburg Binder Design with Disulfides (see the feedback on the Foldit blog for an idea of what we're looking for), we'd like to revisit this design target. The Marburg virus is a cousin of the Ebola virus, and causes a similarly deadly hemorrhagic fever. The largest outbreak, in 2004, had a 90% mortality rate. Understandably, we'd like an anti-viral agent to treat this disease.

Like Ebola, Marburg virus uses a viral surface glycoprotein to grab onto certain human cell membrane proteins and enter human cells. It is our hope that, by blocking the binding site on this glycoprotein that recognizes and binds to human cellular proteins, we can prevent cellular entry and neutralize the virus. To this end, we want you to design a small (37-residue), heavily disulfide cross-linked peptide that can bind to the Marburg glycoprotein. As a starting point, we're giving you a loop taken from a neutarlizing antibody.

This puzzle is similar to Puzzle 1073 in that you will get a substantial bonus for forming up to two disulfides. However, we have added the following:

-- A core existence filter. You'll get a substantial bonus for making a hydrophobic core. (We want compact, nicely-folded peptides, not elongated designs. Please see the Foldit blog for feedback from the last Marburg puzzle, with examples of the sorts of designs that we'd like to see.)

-- A fragment quality filter that penalizes portions of your design that are in unusual conformations that do not resemble those of natural proteins. (Note that this unfortunately scores the entire structure, so all players will see a penalty of 1400 points for parts of the Marburg glycoprotein and for the antibody loop. Because this penalty is the same across the board, it gives no player an advantage or a disadvantage over any other player. Additional penalties can result from having bad loops in the design, though.)

-- A larger penalty for glycine. While glycine is sometimes necessary, too many promote disordered peptides.

-- A larger bonus for proline. Prolines are terrific for making a peptide nice and rigid.

The filters can be a bit slow, so you might want to disable them during design and only re-enable them as needed.
Categories: Design, Overall

Top Groups

2Go Science10,53881
3Anthropic Dreams10,53364
5Beta Folders10,52539

Top Evolvers

Top Soloists

Need this puzzle? Log in to download.  


Enzyme's picture
User offline. Last seen 48 years 5 weeks ago. Offline
Joined: 07/10/2008
Groups: None

1) after deleting and adding segments. the Delete button is now missing from one side of the protein http://fold.it/portal/files/chatimg/irc_104378_1435805368.png

2 and 3) Wiggle lock on unlocked segments hinders manipulation on half of the designable segments http://fold.it/portal/node/993937

v_mulligan's picture
User offline. Last seen 2 years 1 week ago. Offline
Joined: 03/04/2009
Groups: None
Re: issues

1) Sorry about that -- we missed one flag during the puzzle setup, which is causing the deletion issue. Since this is a minor issue, and since players have already been playing this puzzle for a while, we won't repost this puzzle, but it will be corrected in subsequent puzzles. Apologies!

2) The "wiggle lock" problem is actually a family of problems, some of which are merely counter-intuitive but correct behaviour of the Rosetta minimizer (what we call "wiggle" in Foldit), others of which are actual issues. (Rosetta is the protein folding software that underlies Foldit, but which is also used independently to design proteins and to predict protein structure.) In this case, I think it's an actual issue with the way in which we're setting up the Rosetta "foldtree" (the hierarchy of amino acid residues that determines what's fixed/rooted and how children move relative to their parents). We will be able to correct this for the future, but it will take some time.

Thanks for the feedback!

Joined: 09/24/2012
Groups: Go Science
fragment filter on the virus

In all my solutions and the group solutions I've seen, a malus for bad fragment still remain on segments of the virus itself.

I thought the fragment filter checked how "natural" a structure was ?

Or is this a bug? (I would expect the filters would help us to change thinks, thus not applying to locked segments).

v_mulligan's picture
User offline. Last seen 2 years 1 week ago. Offline
Joined: 03/04/2009
Groups: None
Re: fragment filter on the virus

Yes, there was a note about that in the puzzle description. We hope to be able to apply the filter to just the designable regions in a puzzle in the future, but we don't have that feature available to us yet. It's true that it doesn't really make sense to be penalized for odd structural features in the natural protein, but since it's a penalty that applies to everyone uniformly, it doesn't put anyone at a competitive disadvantage.

As for the other part of your question, yes, the fragment filter is intended to tell you the extent to which a structure resembles natural structures. The way in which we do this is by comparing each segment of a design to a database of fragments of carefully-selected "good" structures. Now, this database is not an exhaustive snapshot of all of the protein structures in the Protein Data Bank, for several reasons. First, there are many low-quality or low-resolution structures that we would not want to use. Second, there are lots of proteins that "break the rules", especially with specialized loop conformations that can only exist in very specific contexts, but which are not generally representative of what proteins typically do. We want the fragment filter to be pretty conservative, yielding false negatives (reporting that a design is bad when it might actually be okay) rather than false positives (reporting that a bad design is fine). Although nature seems to know when it can break the rules (as this virus protein has done), we have the most success when we try to keep our designs strictly canonical.

Get Started: Download
  Windows    OSX    Linux  
(10.7 or later)

Are you new to Foldit? Click here.

Are you a student? Click here.

Are you an educator? Click here.
Only search fold.it
Recommend Foldit
User login
Top New Users

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, Microsoft, Adobe, RosettaCommons