You can see the variation in the models solved by NMR:
http://www.rcsb.org/pdb/explore/explore.do?structureId=1NSO

Our hope is that Foldit players can come up with a model that fits the more recent X-ray crystallographic data better than these NMR models from 2003. Then we could use that prediction for molecular replacement (http://en.wikipedia.org/wiki/Molecular_replacement) and solve this monkey virus protein using X-ray crystallography!

This would be an amazing scientific achievement, as we have been unable to use Rosetta to solve this particular structure using molecular replacement, but our lab has been able to do it with other proteins.

So we are giving you all 10 NMR models as starting structures (every time you reset the puzzle it will randomly select one of these 10 structures) and all 10 starts are also available in your Template Reserve in the Alignment Tool (as well as an extended chain conformation).

Even though the sequences for all 11 structures in the Alignment Tool are all identical, threading them will give you very different results. We hope that you can use partial threading to combine different regions from each NMR model.

Remember that by clicking on the 'Score' button on the top right of the Alignment Tool, you can see the name of each model ('model1', 'model2', etc., and 'extend' for the extended chain conformation) so that you can easily differentiate between the 10 different models. You can go back to the alignment score (which is useless for this puzzle since all the sequences are identical) by toggling the 'Name' button:

http://foldit.wikia.com/wiki/Alignment_Tool

" ...

14. Set Partial Threading: To thread a subset of your query sequence, select a region (by clicking, double clicking, or shift clicking to select more letters) in the query sequence, click this button to highlight it and thread [5].

15. Clear Partial Threading: Click on this button to clear the partial threading region you previously selected and return to the entire alignment.

..."

buttons 14 & 15 are the two gray squares above "Name" in the picture in comment #1, according to the illustration at the above link. Thank goodness for Google advanced search ;-)

Thank you for that, we really should move the Alignment Tool explanation to the FAQ or somewhere more visible!

When I restart this puzzle I get one of the 10 models (0,1,...,9) but the start score isn't the same as any of the post-threaded scores for the models in the reserve.

How can I tell which model I have after I randomly reset? The ten possible starting scores (exclusive of "extend") are

A) -2844.163
B) -6889.592
C) -4259.602
D) -2499.298
E) -4012.628
F) -3134.392
G) -3847.071
H) -2166.369
I) -5740.561
J) -470.151

Start scores of the models as they come up in the random resets:

Model 0: -2166.368
Model 1: -4259.601
Model 2: -2844.162
Model 3: -5740.560
Model 4: -470.150
Model 5: -3847.070
Model 6: -2499.297
Model 7: -6889.591
Model 8: -3134.391
Model 9: -4259.601

======================
This is truly ugly. In puzzle 390 each of the Models has a slightly different structure. After you've threaded from one of the models the following source code does fifty puzzle resets (destroying your present position!) and puts the start-score of any resets with the current structures into the Recipe Output. If you're really unlucky you might have to run it twice to get the random hit.
=======================
-- Report Reset of Current Structure destructive v0.01; prints start scores of resets that match
-- current structure (**warning** destructive script doesn't restore current)
local string current_structure={} -- table of present structure codes for all segments
do -- start constructor for table current_structure
for x=1,get_segment_count() do
current_structure[x]=get_ss(x)
end
end -- constructor for table current_structure
for x=1,50 do -- number of resets, these come up randomly
reset_puzzle()
local boolean same_structure=true
for y=1,get_segment_count() do
if current_structure[y] ~= get_ss(y)
then
same_structure=false
end
end
if same_structure
then
print(get_score(true)) -- present game score, including negative scores
end
end

Start scores of the models as they come up in the random resets:

Model 0: -2166.368
Model 1: -4012.627 **corrected**
Model 2: -2844.162
Model 3: -5740.560
Model 4: -470.150
Model 5: -3847.070
Model 6: -2499.297
Model 7: -6889.591
Model 8: -3134.391
Model 9: -4259.601

I nominate this protein as the inaugural entry in the ugly proteins gallery: this might be the companion blog to Mike Tyka's beautiful protein blog. See http://beautifulproteins.blogspot.com/

gstart={
{0, -2166.368},
{1, -4012.627},
{2, -2844.162},
{3, -5740.560},
{4, -470.150},
{5, -3847.070},
{6, -2499.297},
{8, -3134.391},
{9, -4259.601}
}
gss={}
for j=1,get_segment_count() do
gss[j]={[L]=0,[H]=0,[E]=0,[M]=0}
end
local startsFound=0
repeat
reset_puzzle()
local i = 0
repeat
i=i+1
until (gstart[i]%1)==(get_score(true)%1)
if gstart[i][3] == nil then
startsFound=startsFound+1
gstart[i][3]='found'
for j=1,get_segment_count() do
ss=get_ss(j)
gss[j][ss]=gss[j][ss] + 1
end
until startsFound=10
-- Now if my code is correct
-- you have a vote on secondary structure.
-- you don't need to have the scores preloaded
-- but what the heck. Do you want a vote on
-- segment distances as well? You might be
-- able to band to those distances.

You tell us the NMR data does not fit perfectly with the X-ray data, and give us the NMR templates and graph, but no information about the X-ray data. I believe, depending on the resolution in Angstroms of the X-ray data, it is possible to determine secondary structure, but that loops tend to be undetermined. Can you tell us the resolution of the X-ray data and what secondary structure information was extracted from it, or any other information you can give us to help fold this mysterious protein?

In general, secondary structure composition can't be determined from unphased crystal data. There are some exceptions with all-helical proteins (http://www.nature.com/nmeth/journal/v6/n9/full/nmeth.1365.html)

What we are hoping will work with Foldit is what was done previously using Rosetta: (http://www.nature.com/nature/journal/v450/n7167/abs/nature06249.html).

Most likely, the topology & secondary structure of the protein is pretty similar in the X-ray structure as in the NMR structure. Molecular replacement generally fails above 2A RMS so the target may not be very far from the NMR ensemble. That said, there may be some reorganization due to crystal contacts ... but sadly we really don't know.

Sorry about that, you can download the pdf from the Baker Lab webpage:

http://depts.washington.edu/bakerpg/drupal/node/115

You'll find the Foldit nature paper there as well, if you didn't get a chance to read it:

http://depts.washington.edu/bakerpg/drupal/node/16

If you want to look at some X-ray structures that might be familiar to you, you can go to the CASP9 target page:
http://predictioncenter.org/casp9/targetlist.cgi

and click on the blue PDB code on the right for any of the targets that have been solved using X-ray crystallography (those are denoted by the 4th column on the CASP page saying "X-RAY")