Tool suggestions

Case number:699969-2001025
Opened by:Brian the Prion
Opened on:Friday, August 7, 2015 - 11:14
Last modified:Saturday, August 8, 2015 - 01:32

My submission has triggered the spam filter, so I here is an external link to my message:

(Fri, 08/07/2015 - 11:14  |  2 comments)

brow42's picture
User offline. Last seen 1 year 35 weeks ago. Offline
Joined: 09/19/2011
Groups: None

Text of the Link

7 August 2015

I started on FoldIt yesterday, and have been working on Beginner Puzzle (<150): Easy Mini Freestyle, where I am less that 60 behind the highest scorer, and 1120: Unsolved De-novo Freestyle 54, where I have about half of the highest score. I have just started on 1123: Unsolved De-novo Freestyle 54: Predicted Contacts.

These puzzles starts with the protein laid out in a straight line.

I have been working with Shake and the various Wiggle tools, and I have also used Idealize, Ideal SS, Rebuild and rubber bands. However, none of these tools allow a precise manipulation of the protein string. If I have a precise idea about how I imagine two parts of the structure to be related to each other, I cannot move them directly into that position.

I understand that the advantage of using humans to provide suggestions on how proteins are folded is that humans can use insight in a way that a data-crunching computer algorithm cannot. It appears that the players with the best scores have an implicit understanding of protein structures, gained (no doubt) from hours of tweaking proteins into place.

I have some ideas for additional ideas that I believe would make it easy to create the overall shape of the protein, so that the Wiggle tools can be used to optimize a well-considered starting point. Perhaps there is already a way to do the things I imagine. If so, I would be grateful for guidance on how to go about it. Perhaps there are already scripts which make some of these ideas possible.

Hiding sections of the protein string. I imagining labelling different points on the protein, and being able to toggle the visibility of any section between two points. This would mean that you could start in the centre and work both ways, or hide the outer sections while you tweak the area in the centre, or position a particular feature inside an electron density mesh. Hidden parts would not generate any clashes.
The Rebuild method appears to act just on the chosen section. The sections at either end remain in a fixed position. As far as I can tell, it does not track the best value that it encountered as it generates random paths between two end points. Would it be possible to...?
Provide a graph showing the Rebuild method's different mutations. This could be score against time, like the Undo graph. Dragging the mouse along the graph could update the view to show the mutation that was suggested at that point in time.
Provide the option of fixing only one end of the Rebuild section, so that the other end is free to move as the Rebuild section changes shape.
the protein that get folded back into the part you are working on.
Alignment axis tool for helices. I am imagining that the axis of each helical section could be made visible, and that it could be possible to define a line using two points in space. The axis of the helix could then be aligned with the line, and the helix could be rotated around its centre and translated along it. If a Wiggle tool is used, and the axis of the helix moves, then the line defined previously could be adjusted so that it is still at the centre of the helix. This would be particularly useful in situations where the electron density data makes the position of a helix clear.
Ability to curve (a section of) a long helix into a toroid shape and to twist a helix around the its own spiral, like straightening your stockings.
Fit protein to window. Sometimes an operation like Ideal SS will move the entire protein outside the view frustrum, and it can take a while to find it again. A button that moves the camera to recentre the protein and make it fit the window would be very useful
Pin a given point to a given pixel on the screen. While this feature is active, all zooms and rotations would be around one fixed point in the protein.
Lock the protein while adding bands. Often I will set up the view to make the points to which I want to attach a band visible, but the model will rotate as I am dragging and the end point will become inaccessible. An alternative would be to use Shift-click to define the start and end points, with no dragging required. This would also allow dragging to scroll, rotate and zoom before the second Shift-click.

Susume's picture
User offline. Last seen 1 day 5 hours ago. Offline
Joined: 10/02/2011

Many of your concerns are addressed by tools and techniques that already exist. I suggest practicing with cut points and the move tool first of all, to enable you to put parts exactly where you want them in space. These are covered (slightly) in tutorials 5-2 and 6-2. There is a learning curve associated with closing cut points, so give yourself time to learn it and feel free to ask questions in chat.

Some of your camera wishes have keyboard shortcuts; see
Typing Q with mouse over background: brings protein to center of view
Q with mouse over protein backbone: pins view center to that point in the protein
Shift Q with mouse over backbone: centers on that point, zooms, and fades foreground and background to make the plane of that point visible. Q returns to normal view.
Ctrl-Shift drag up/down fades background
Ctrl-Alt drag up/down erases foreground

Regarding Rebuild, the purple positions that fly by quickly are positions that get rejected because the endpoints cannot be rejoined after that rebuild. Each rebuild position that is not rejected shows up as a point in the undo graph. You can work through them from right to left to find one that scores well after a shake and wiggle. Hovering over a point on the undo graph will show the score of that position. Cut points will also allow the end of your rebuild area to move freely.

To form a perfectly straight helix or sheet, right click it and choose Ideal SS. Putting cut points around it first will keep the rest of the protein from moving when you do this. To advance the amino acids one place along the spiral of a helix, change the helix to sheet and use the tweak tool to advance the curled sheet one position. Then change it back to helix. If you want to straighten a helix or sheet a little but not all the way, use the tweak tool and click on the purple sphere instead of the arrows.

Feel free to ask more questions in chat; there are usually friendly people around to give pointers.


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