Does Foldit penalize leaving high density parts of the ED cloud unoccupied?

Case number:699969-2012747
Topic:General
Opened by:jeff101
Status:Open
Type:Question
Opened on:Wednesday, February 9, 2022 - 00:33
Last modified:Wednesday, February 9, 2022 - 21:00

Some solutions for Puzzle 2100 lead me to 
questions about how ED Puzzles are scored. 

If you compare the initial conformation (IC)
with score 14555.167 with the saves below:

Je2 22468.287 med 408pm 2/2/22 
NG  22517.873 med 419pm 2/2/22

You'll see that in IC, arg57 lies in a region 
with some density in the ED cloud. Meanwhile, 
Je2 and NG above have arg57 in a different 
position and leave empty the part of the ED
cloud where IC had arg57. Does it make sense
to leave parts of the ED cloud with high 
density unoccupied by protein?

I think ED Puzzles are scored so that it is 
best to put high density parts of the protein
(like aromatic sidechains and helices) in high
density parts of the ED cloud. It is even ok
to have high density parts of the protein cross
through low density parts of the ED cloud if
those parts of the protein are quite mobile.
What happens if a region of the ED cloud with
high density is left unoccupied by protein?
Does the puzzle score encourage protein to fill
that part of the ED cloud? Should the Foldit 
score penalize leaving high density parts of 
the ED cloud unoccupied? Could the unoccupied
but high density parts of the ED cloud be for 
nearby proteins that we aren't modeling? 

Just curious.

Thanks for reading.

(Wed, 02/09/2022 - 00:33  |  2 comments)


horowsah's picture
User offline. Last seen 2 weeks 3 days ago. Offline
Joined: 03/17/2016
Groups: DU Proteins S20

In theory, yes, but there are some approximations made such that there are imperfections that can occur. These tradeoffs are needed to make it fast enough to work in scoring.

horowsah's picture
User offline. Last seen 2 weeks 3 days ago. Offline
Joined: 03/17/2016
Groups: DU Proteins S20

I looked into this a bit more closely- it looks like that extra density was an artifact, which does happen when dealing with experimental data that's imperfect. So fitting that data would presumably make other things in the model worse, hence leading to a lower score. Not an easy case to spot whether to fit or not to fit that bit of cloud.

Sitemap

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, Amazon, Microsoft, Adobe, Boehringer Ingelheim, RosettaCommons