How confident are you that the locked fragments are placed correctly? They don't seem to match the density super well.
The N-terminus is in the general vicinity of the cloud.
It is probably closer to the little isolated clouds of density that are at the edge of that region (they look like 4 islands breaking off).
I'm sorry I can't post a photo (thank goodness the new website is coming!)
We just didn't want to have you search the entire cloud for the N-terminal end of the density.
In Puzzle 1952, we have several small
locked fragments from segments 1-67
and one long continuous unlocked
fragment from segments 68-154.
None of the 154 segments are mutable,
but many have more than one rotamer,
so wiggling sidechains or shaking for
both segments 1-67 and segments 68-154
should affect the score. Nevertheless,
rebuild, remix, or wiggle backbone should
only alter the score for segments 68-154.
Viewing with cpk, stick, all hydrogens, & all sidechains
shows that residue 68 (e=glutamate) ends with NH3
while residue 154 (l=leucine) ends with COO.
Does this mean the N-terminal of our unlocked fragment
is at residue 68 while the C-terminal is at residue 154?
Does Puzzle 1952 have any constraints holding the
N-terminal of segments 68-154 at its desired location?
Can you please describe the N-terminal's desired
location using distances between glu68's NH3 nitrogen
and several nearby backbone atoms on the locked
segments in Puzzle 1952?
Finally, does the ED Cloud for Puzzle 1952 contain
any other protein fragments besides the ones given
to us as segments 1-154 in this puzzle?
Now you can view the entire complex from the PDB if you like!
Thanks for all your answers so far about
Are there certain N-terminal atoms in this
puzzle's starting structure that are very
close to their desired positions in the
ED Cloud? Perhaps we could use zero-length
bands from these atoms to space to keep
these atoms in their starting positions.
If so, which N-terminal atoms would you
fix in place with respect to the ED Cloud?
but hopefully this view helps: https://i.imgur.com/lwSb2lK.png
Again, keep in mind this could be incorrect... especially since the ends of a protein chain tend to flop all over the place! But it's at least probably in that general area.
I am away for 10 months or so.......it feels like home to see the old issues have still to be addressed :)
but I guess that is progress?
Or are you and alcor29 referring to something else?
I got this SS from psipred (translated into jpred format just 'cause I like that better; H=helix, E=sheet):
This puzzle kind of strains AA Edit/SS Edit/print protein, since it has protein chains A through I. The dialogs in these recipes may be too tall to fit on some screens. SS Edit is probably the best bet of the three size-wise. I'll work on scrollable dialogs to fix that, but in the mean time here's the takeaway.
Chain I is the foldable part, while chains A through H are anchored somewhere in the cloud.
The chain I AA sequence is:
Trusty Jpred finds a number of exact PDB matches.
Chain I in the Foldit puzzle exactly matches chain E in these PDB entries:
5ayw 5d0o 5d0q 5d0q 5ekq
5ljo 6lyq 6lyr 6lys 6lyu
6sn9 6sob 6soj 6v05
As is often the case, some interpolation is necessary to get the numbering schemes to line up.
In Foldit, the foldable chain I starts at segment 68, and runs to segment 154, for a length of 87 segments.
Looking at 6v05 in JMol or the PDB viewer of your choice, the first five residues are missing for chain E. The model starts with arginine as residue 29, which corresponds to segment 73 in Foldit.
So for 6v05 at least, the rule is to add 44 segments to what you see in your PDB viewer to get the equivalent Foldit segment.
No guarantees on my work, but here's the secondary structure for chain E from PDB 6v05:
Given the existence of PDBs that are exact matches, to what extent are people allowed to consult publicly available external sources like 6v05?
According to https://fold.it/portal/communityrules#cheating , copying is allowed for primary and secondary structures, but not tertiary and quaternary structures. But what about "Level 2.5" approaches, such as noting that two specific beta strands go together (while NOT going as far as to attempt a full duplication down to the residue / angle level)?
Also, could there be an option to designate an existing structure as "evolved from external sources," i.e. voluntarily exclude it from soloist competition for the sake for fairness (due to usage of the "Level 2.5" approach described above, etc.)? Alternatively, could the known native be provided as a "global evolver solution" (i.e. allow everyone to load in directly into Foldit) order to level the playing field?
In my opinion, Foldit team should give an SS prediction for this kind of difficult puzzle. This would help us to find an approximate match in the density, refining later on.
As LociOiling discovered, the chain we gave you to fold into the density is 6LYU chain E:
which is the shortest chain from the 6LYU complex: https://www.rcsb.org/structure/6LYU
(You can see that we fixed segments of the other chains that are closest to chain E)
We want to emphasize that we believe the deposited structure has errors, because you can see how poor the PDB validation scores are (particularly the clash scores on page 2):
Lastly, we hope to have future ED puzzles that will be blind (where the native will not be available) so we want to encourage you to practice folding without the native. This, of course, is difficult when anyone can look up the PDB.
Perhaps we can suggest folding it at first with just the given density, and then looking up the reference structure in the last few days of the puzzle?
It's been seven years since I've done an ED puzzle, so I don't know how they've evolved. Please forgive any ignorance.
The puzzle supposedly has locked fragments inside the density cloud. When I moved my design inside the cloud, my designed structure stayed pretty much the same, but the fragments exploded in all directions and landed a full monitor screen away from the cloud. How is that possible, if they are locked?
Also, just rotating my design along an axis, while nowhere near the cloud, dropped my score by thousands of points. How is that possible, when nothing was done to my design's structure, and it was too far from the density cloud to matter?
In addition to the ED Cloud you can see,
many of these puzzles have invisible clouds
nearby. It is like a periodic lattice of
ED Clouds in all directions with some empty
space between them. If a segment of your
protein is in the empty space, it gets a
zero or negative Density subscore. If a
segment of your protein is in an ED Cloud
(either the visible one or one of the many
invisible clouds), it gets a positive
It is possible to have your protein be
straddling more than one ED Cloud. Then
you will have some regions with zero or
negative density (between the clouds)
and others with positive density (within
Below is a Feedback with a diagram of a
periodic lattice of ED Clouds.
I have written many other Feedbacks about
ED Puzzles, some as listed below:
Indeed, this is the case with the ED cloud in Foldit (as well as in many other density programs)
The only workaround would be for us to pad the map with a buffer of zero density, but this would sadly increase the size of the map in memory... which we obviously don't want to do! (as demonstrated by your feedback: https://fold.it/portal/node/2011161#comment-44012, we don't want to increase that even more!)
Boots, your second question is answered above, but we are stumped by your first question.
Has anyone else run into this bug?
Boots, do you have a screenshot or scientist save of this explosion?
I was just looking into the issue the Boots reported.
In mine, segments 5-9 are now a long distance outside the cloud.
Uploaded for scientists as "segments 5-9 ejected".
Not sure how to fix it, I guess I can band them back to where they started.
I was so taken aback at "locked" segments flying away that I undid the action and performed other actions, erasing the one that caused the explosion. I simply moved my protein inside the cloud and wiggled.
I shared with scientists several solutions
before and after segments 5-9 got ejected
from the cloud. In each case, I was using
ci=1 at low wiggle power and tried wiggle
backbone. This gave a drastic change in
the protein's shape, raised the score,
and ejected segments 5-9 from the cloud.
Starting from the solution Je2b -91877.294,
wiggle backbone ejected segments 5-9 and
got to -10266.665 before I quit it. After
that, I did shake and then wiggle sidechains
to reach -7767.930. That's when I noticed
segments 5-9 had ejected. Then I used undo
to find which step did the actual ejection.
Starting from the solution Je2b -91842.788,
wiggle backbone ejected segments 5-9 and
got to -7087.026 before I quit it. Then I
used undo to restore -91842.788 and tried
wiggle backbone again. This time it ejected
segments 5-9 and got to -7086.563 before I
Say a player puts some of 1952's protein fragments
in a visible ED cloud and other fragments in an
identical but invisible ED cloud. Both ED clouds
contain the same pockets A-D. Say fragments
a-d belong in ED cloud pockets A-D respectively,
but the player puts fragments a-c in pockets
A-C in one ED cloud and fragment d in pocket
A in the other ED cloud. Being in distinct
but identical ED clouds, fragments a & d would
be too far apart to clash with each other, but
they might both fit very well into pocket A.
This would give good density subscores for
all of the fragments, but it would neglect
the score penalty for putting both fragments
a & d into ED cloud pocket A, a pocket just
large enough to hold either fragment a or d
but not both fragments a & d simultaneously.
Maybe the scenario above can explain why
Boots' exploded protein scored so well in
The exploded version was not my highest scoring. My current score is completely and far away from the density cloud. No amount of manipulation inside the cloud will give me the score I had prior to moving it inside.
We believe that we have found what caused this.
We plan to repost this same protein again soon, this time starting from the PDB model (so that you can hopefully improve on it :-)
Thanks again and our apologies for the trouble.