New LayerDesign filter is not usable

Case number:845813-993879
Topic:Game: Tools
Opened by:mbinfield
Opened on:Wednesday, November 14, 2012 - 02:27
Last modified:Thursday, November 15, 2012 - 22:40

I appreciate the goal of the new LayerDesign filter for mutating, but the implementation doesn't work. It's not just preventing unrealistic designs, it's inhibiting any serious manual design work.

When constructing a design by hand, there are a lot of intermediate steps that aren't intended to be stable, well-folded proteins. Unfortunately the new filter has no way to account for that. So, I can fold up the protein into a stable shape I like, hit mutate, and accept whatever side chains appear. That level of design still works fine. But more sophisticated work, like creating a disulfide bond or getting good hydrophobic packing between aromatic rings, has become much more difficult, because I have to first get the protein into a configuration that will allow me to place the side chain, without being able to see the side chain. The filter seems to be very picky about allowing placement of hydrophobics. In fact, so far I haven't been able create a disulfide bond on the new freestyle design puzzle, because no matter how I fold the protein, I can't get the option to create a cystein.

One possible fix would be to allow us to turn the filter on and off. Alternatively, it could be made somewhat less discriminating, so that the protein doesn't need to be in such a specific configuration before allowing a side chain to be placed. Or maybe putting restrictions on the mutate tool is the wrong approach entirely (that's where I'm leaning so far.) Regardless, something needs to change.

I'm ranking this as Priority 1 because as far as I'm concerned, manual design will be broken until LayerDesign is fixed or removed.

(Wed, 11/14/2012 - 02:27  |  4 comments)

Joined: 08/30/2011

not just manual design will be broken, in my 3 designs the hydrophobics are quite rare. without any manual mutating.

tamirh's picture
User offline. Last seen 8 years 49 weeks ago. Offline
Joined: 05/11/2012

Thanks for the input. We definitely don't want to make it too difficult to manually fold when the layer design filter is on. We have a layer design score filter. It uses s the same basic algorithm but instead of preventing mutation, it will just dock the score. This way you are free to go into those intermediate bad configurations. You will also have an easy way to turn on the visualization that will show you exactly which residues are considered bad.

We shied and away from the on/off option because it would create a situation where you
may have a very good score but it is an invalid score because the layer design filter is off. When you turn the filter on and fix the residues the protein may no longer be stable.

Joined: 08/24/2011
Groups: Go Science

Thanks for the response. Based on my experience so far, I would definitely recommend focusing on the score-based filter rather than preventing mutation.

It occurred to me later that preventing mutation might be more appropriate for more constrained puzzles like the flu series or the WNT puzzle. It's just not a viable option for freestyle design.

spdenne's picture
User offline. Last seen 7 weeks 1 day ago. Offline
Joined: 10/01/2011
Groups: Void Crushers

I'm finding it really hard to change designs too. For example, I'm currently unable to change a serine (which has 2 bonds to a helix) to a glycine.


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