More precise manipulating of segments

Case number:845813-990716
Topic:Game: Tools
Opened by:Nas Brenton
Opened on:Wednesday, October 5, 2011 - 10:46
Last modified:Wednesday, October 5, 2011 - 22:36

Pulling and bending segments is too inaccurate. If you truly want people to be able to fold proteins by hand, we need to be able to bend, shape and twist each individual segment of the protein how we like, rather than pulling on something and affecting changes in the whole protein. Otherwise all fold it is really doing is relying on people writing scripts and spending time running them, which really is not people powered folding. We may as well be using folding at home instead.

(Wed, 10/05/2011 - 10:46  |  2 comments)

Joined: 09/20/2011
Groups: None


Foldit needs a mechanism (both tool and lua functions) by which to precisely and independently manipulate the orientation of each amino acid segment and sidechain. This is critical for both tricky manual adjustments as well as for scripts.

For example, when playing Puzzle 459 I found that as my score climbed higher I could "see" how the protein was "supposed" to look, but (even with multiple days of tedious banding and selective freezing) I was unable to get the protein to assume the desired shape.

As another example, suppose I want to write a new script that utilizes a genetic algorithm. Is a method like banding capable of producing all possible valid shapes of the protein? What does a "wiggle" or a "shake" really do, for that matter? Questions like these don't need an answer if I can access the fundamental structure of the protein directly via lua function calls! Then, instead of guessing, I can be sure to write my algorithm in such a way as to guarantee theoretical soundness.

A final suggestion concerning this is to implement a new tool called "shape". "Shape" would allow the user to select one or more segments (in a row, index-wise) and copy their orientations. Then, the user would select a different (but equally sized) row of segment(s) and "shape" would attempt to re-orientate this second selection to match the first (either matching it exactly or otherwise finding and applying the closest valid form).

The developers need to understand that advances to the current protein folding methodologies will likely not occur (with any real significance) until they extend Foldit to include these capabilities!

Let's hear some real feedback on this issue, please!

Mark-'s picture
User offline. Last seen 33 weeks 1 day ago. Offline
Joined: 08/05/2009
Groups: Contenders

I don't think the problem is more accurate backbone manipulation. Foldit used to work so that when a part of the backbone was close enough to a 'correct' shape the wiggle would take care of most of the minor adjustments needed to find that area's lowest energy. As it stands now, scripts are almost essential to go through all the posssible backbone configurations to find one close enough to achieve the lowest energy after wiggling. As it stands it is just wasting time and resources and pushing those of us who prefer to fold manually out of the game. Hence the reason I play so little now and don't even bother with some puzzles at all.

Can someone please try and find those old settings, or at least experiment with some new ones in beta.

I have no interest in running scripts for hours on end. If I wanted to do that I would dedicate all my cores to Rosetta.

... and yes, I do feel strongly about it :)



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