Show bad fragments (without score filter) in non-design puzzles

Case number:845833-997305
Topic:Game: Display
Opened by:Susume
Opened on:Sunday, March 23, 2014 - 17:20
Last modified:Thursday, March 27, 2014 - 18:43

It would be useful in non-design puzzles to have a view option that would light up unrealistic fragments so we could fix them. Don't penalize the score based on fragments; just show us where they are. Since this would be slow, only test the fragments if the view option is checked. Just like show sidechains, the view menu could say "(slow)" next to the option.

(Sun, 03/23/2014 - 17:20  |  2 comments)

jflat06's picture
User offline. Last seen 19 hours 16 min ago. Offline
Joined: 09/29/2010
Groups: Window Group

The problem with this is that we only use a subset of realistic fragments for the visualization. It is intentionally overly restrictive because in design we don't care that we are potentially ruling out a 'good' fragment. We'd rather be safe and restrict you to only the best fragments, because this gives your designs a much higher chance of actually folding.

But for prediction, this becomes a problem. The native fold of a particular protein might actually have one of those 'good' fragments that we've ruled out, so biasing you away from that fragment effectively rules out the correct solution.

This is also the same reason why loop hash (Remix) is not enabled in prediction puzzles.

Susume's picture
User offline. Last seen 14 hours 31 min ago. Offline
Joined: 10/02/2011

Thank you for the explanation.


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