I posted some questions about Electron Density Puzzles in Feedback at http://fold.it/portal/node/2001368
If you know the answers, please post them there. Thanks!
In the spirit of fairness for this Foldit vs. UMich Electron Density Challenge, the Bardwell Lab at the University of Michigan is kindly sharing with us additional information that they provide their students.
Here are some hints they gave out in class today:
"In order to help place the sequence, it helps to look for groups of large bulky aromatics, like tryptophans, phenylalanines, and tyrosines in the sequence, as these will be the most identifiable spots in the structure."
They also pointed their students to this useful picture:
In addition, I've seen in their paper on 2014:
"Horowitz positioned a single amino acid on the map to show where the protein chain started."
Did they also get this information for puzzle 1152? If it's the case, could we receive the same info ?
This year, the students were not given the position of any amino acids in the protein.
They have only been given the sequence and the exact same secondary structure predictions that were provided on your extended chain.
Any additional information (not related to Coot) that is given to the students by their instructors will be posted here. This is a class after all, so the students will undoubtably ask questions!
I know you already voiced your request for a repeat of their previous experiment:
but Foldit players showed in the subsequent Foldit puzzle (http://fold.it/portal/node/2000180) that you could solve these ED problems without that information.
This experiment is much more realistic (in how Foldit could help with protein structure determination) because it is unlikely that we would know that information for an unsolved case.
The UMich class has recently spent some time discussing the topic of sidechain flexibility. We want to make sure Foldit players are aware that some long, flexible sidechains will not fit into any density.
You may have already had some trouble finding density for long sidechains like those of LYS, ARG, GLU, or GLN. When the protein is in solution, these residues are particularly flexible, and shift frequently between different conformations. Often, and especially on the surface of the protein, these residues can retain their flexibility even in the crystal form, when most of the other residues have been locked down into a single, rigid conformation.
The uniformity and rigidity of molecules in a crystal is key to obtaining high-resolution electron density. A flexible, nonuniform residue in a crystal only gives a noisy signal; as a result, these flexible residues appear to have no density in the final density map.
I think it's useful as a technique for hand folding first steps of EDs in Foldit:
"The trained crystallographers reported using the following approach to the puzzle, which corresponds well with what the instructor observed with many of the undergraduate students. First, they looked for large density blobs that might correspond to large aromatic side chains such as Trp, Tyr or Phe. Working forwards and backwards from the Trp–Phe–Val–Asn sequence proved particularly useful. Modelling in a few of these large residues led to the assignment of density to sequence location. The direction of the polypeptide chain was reversed on a few occasions, but was fixed by looking at the carbonyl density. The Find Secondary Structure tool in Coot was used, especially for regions where the density was poor. Real Space Refine Zone was used with the refinement weight set to 20 or 10, based on the instructor’s suggestion for building in an unrefined map. Regions where the density was very poor and decisions had to be made about whether to keep trying to build or not proved to be the hardest part of the task. The trained crystallographers reported that at first they did build in these sections of poor electron density. However, when they realized the extent of the guessing involved, they subsequently removed most of the model in these areas. After modelling in the residues, the trained crystallographers used the validation tools in Coot, including Ramachandran plot, Rotamer analysis and Density Fit analysis, which flagged areas with poor geometry. They also ran the structure through MolProbity, which gave similar results to the Coot validation tools. Finally, the crystallographers fixed problem areas as best as possible with the Coot modelling tools, such as Flip Peptide, Rotamers, Regularize Zone and Real Space Refine Zone. When asked to describe the difficulty level of this assignment, the trained crystallographers rated it as somewhat difficult (on a scale of: very difficult, somewhat difficult, neither easy nor difficult, somewhat easy and very easy)."
Susume has made a new video about ED puzzles.
See the link below for more info:
Thanks again, Susume!
I don't know if it makes sense, but would it be possible to share the best result (I suppose the published one?) as an "expired" 1152b puzzle for us to evaluate it and see our differences?
(and who knows, even try to further evolve it :)
It can be useful for learning.