"the wiggle all"

Case number:699969-2011617
Topic:General
Opened by:sky star
Status:Open
Type:Suggestion
Opened on:Thursday, May 27, 2021 - 12:42
Last modified:Monday, May 31, 2021 - 01:16

it starts with the same parameter ,cause the result scores same,because the same algorithm,

(Thu, 05/27/2021 - 12:42  |  1 comment)


LociOiling's picture
User offline. Last seen 8 hours 52 min ago. Offline
Joined: 12/27/2012
Groups: Beta Folders

Not always. A wiggle with the same parameters and the same starting point may produce different results.

I tried a simple test with this recipe:

structure.WiggleAll ( 10, true, true )

Using puzzle 1999, I started with a solution scoring 8163. This solution had only been worked on low wiggle power. In most cases, wiggling such a solution at medium power will produce a gain.

I ran the same test 6 times. The results were:

  • 8419.206 (2 tests)
  • 8320.128 (1 test)
  • 8318.740 (2 tests)
  • 8318.148 (1 test)

This solution started with a Buried Unsats penalty of -800 points, corresponding to 13 buried polar atoms.

The high-scoring wiggles managed to reduce that to -700 points, or 12 buried polars. There were also small changes in the DDG, Contact Surface, and Interaction Energy objectives.

The Buried Unsats (BUNs) objective is very sensitive. Just a small change in the position of one atom can add or subtract 100 points.

Also, the computer I used is fairly busy, running at about 90% CPU, as reported by Process Explorer on Windows.

In general, it's difficult to get the same results in Foldit doing the same thing twice. This is especially true for recipes which are more than one command long, which often run for hours or even days. Sometimes the differences are small, but other times, they are hundreds of points.

For reference, see structure.WiggleAll on the wiki.

Sitemap

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