Which strategy for mutate? Mutate All tool versus scripts/hand

Case number:845813-1998756
Topic:Game: Tools
Opened by:Bruno Kestemont
Opened on:Saturday, November 22, 2014 - 11:29
Last modified:Wednesday, November 26, 2014 - 11:49

I read from the Wiki that the mutate tool does not try all possible mutation, but it uses a "Monte carlo" algorithm in order to save time (see http://www.pnas.org/content/97/19/10383.full).

1) Is it correct that when mutating only one segment, all possible AAs are tested and Monte Carlo does not apply?

2) What happens with the position of the side chains after mutating? Is a shake still necessary? or does the mutate function also look for the best sidechain position?

3) Does the mutate function only test the "preferred" AAs for a specific SS? For example, it would not try an AA that likes sheet in a helix. Or does it test anything just based on a statistical algorithm with pairs of any AAs?

And now the implications for the design strategy.

4) Just after the SS first hand design, does it make sense not to mutate all, but instead, mutating manually (or with recipes) to the best desired AAs (like hydrophobic, or helix preferred where we want an helix) instead of to let the mutate tool do the work?

Many questions. Thanks to anybody who can help.

(Sat, 11/22/2014 - 11:29  |  3 comments)

spmm's picture
User offline. Last seen 8 weeks 3 days ago. Offline
Joined: 08/05/2010
Groups: Void Crushers

Interesting questions Bruno - I am particularly interested in the apparent 'rotation' of the side chains during mutate

Joined: 09/24/2012
Groups: Go Science

I noticed that te mutate all function does not take the filters into account. On 1018, mutate all can result in a lower score because of the filters.

xiando's picture
User offline. Last seen 7 years 15 weeks ago. Offline
Joined: 10/19/2008
Groups: Oma Gawd

Well, I'm glad I read this one. I recently returned to Foldit and was a bit puzzled at the behavior of mutate (never used it before), since I got better scores by handpicking the individual side chains, when futzing with the beginner puzzle "docking" and eventually gave up due to confusion over the )for me) new tools.


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