P1939 Linker new CMS bonus uncapped

Case number:671071-2011036
Topic:Game: Other
Opened by:John McLeod
Opened on:Thursday, January 7, 2021 - 13:19
Last modified:Thursday, January 14, 2021 - 01:27

Main group is I believe equivalent

In new Puzzle 1939 Linker, after about 8 hours running Mutate Combo 1.4 for Layers new ContactMolecularSurface Score +3564.47 points (and Secondary Structures at -300 points). This pose has just been Shared with Scientists. Public Recipe here:

Player ichwilldiesennamen earlier noted this in the Puzzle 1939 comments here:

(Thu, 01/07/2021 - 13:19  |  3 comments)

ZeroLeak7's picture
User offline. Last seen 4 weeks 5 days ago. Offline
Joined: 01/25/2012
Groups: Go Science

yes, the CMS produce wrong AA's, if you mutate!

I have everything documented in Discord:


Joined: 05/19/2009
Groups: Contenders

Thank you for documenting this bug.
Please report and document bugs in one place only, here in feedback, not in Discord.

Joined: 12/06/2008
Groups: Contenders

Most of the top scoring solutions for 1939 were heavily saturated with tryptophans. This indicates a scoring bias towards larger AAs, which are not necessarily the best ones.

I had four tryptophans in my five-helix solution for 1939, all located on the outside of my helices (and flagged as exposed) and not interacting with any other sidechains or either of the target proteins. When I substituted more appropriate sidechains (glutamate, lysine, arginine), my score dropped hundreds of points. That didn't happen before the CMS metric was in play, further lending weight to the idea of a faulty scoring mechanism for CMS.


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