Official CASP results and Exciting Upcoming Puzzle

[beta_helix]

We just got back from the CASP9 meeting where the organizers and assessors released all the results.

Foldit was group 170 and you can see the results for the non-refinement categories here:
http://predictioncenter.org/casp9/results.cgi?groups_id=245&submit=Submit
(This shows all the Foldit predictions for only Model 1, which was always the top scoring Foldit prediction for each puzzle. It only shows a few CASP targets at a time, you have to click on "T0575 - T0604" at the top to see those results, for example)
The results for all 5 submitted Foldit models is here:
http://predictioncenter.org/casp9/results.cgi?view=targets&model=all&groups_id=245&tr_type=all&dm_class=all

All the results for the Refinement category are here:
http://predictioncenter.org/casp9/refinement_mdplot.cgi

You can see that Foldit got the best (closest to the native solved structure) predictions for 2 Refinement targets:
http://predictioncenter.org/casp9/refinement_mdplot.cgi?view=tables&target=TR614
http://predictioncenter.org/casp9/refinement_mdplot.cgi?view=tables&target=TR624
which corresponds to CASP9 Refinement Puzzle 8 & CASP9 Refinement Puzzle 10 respectively.

We will post more details about this (as well as how close Foldit players got to the native in the University Folding Challenge, since we all saw who got the highest score!) soon, but we wanted to let you know about the webpage for those of you who want to dig through all the results.
Thanks to Madde's CASP page, you can see who submitted each of the 5 Foldit models for each target: http://de.foldit.wikia.com/wiki/CASP#CASP9-Puzzles

ALSO WE WANTED TO LET YOU KNOW ABOUT AN EXCITING UPCOMING PUZZLE:

We have a new puzzle that is currently unsolved by protein crystallographers. We have a set of protein models determined by NMR, and these don't currently fit perfectly with the X-ray crystallographic data. We are giving you 10 models as starting structures, and we will test if you can find solutions that fit the crystal data better. We'll post more details about the problem and its broad scientific importance when the puzzle is released.

[Deleted user]

Is there a simple scoreboard that says how each team placed on each puzzle?

Or - how many firsts, seconds, thirds etc. did we get?

I can't even tell from that site if we kicked butt or were simply also-rans.

[infjamc]

Here are the limited information I was able to extract regarding the non-refinement puzzles:

The first link will lead to the results page of the CASP9 website. For each puzzle, clicking on the graph with orange lines will give you the GDT summary plot, one of the measures of how closely the predicted structure matches the native (you want the trajectory to be as close to the lower right corner as possible). Clicking on the graph with green, yellow, and red shades with black lines will give you the alignment summary bar graph; again, a higher alignment score (out of 100) is better.

Overall, Foldit solutions tend to rank in the middle of the pack, with the worse relative performances mostly coming out of puzzles where the template is already very good (for which automatic methods seem to work better). Several notable exceptions exist, however:

• T0538-D1 (the five Foldit solutions were ranked #2, 3, 9, 10, and 11 out of 412!)
• T0544-D1, T0553-D1, T0561-D1, and T0568-D1 (Foldit did above average on a puzzle that proved to be difficult for most groups)
• T0579 (most groups had trouble with this one; the five Foldit solutions were ranked #7, 15, 24, 42, and 113 out of 594. Similar results were seen in the D1 variant.)
• T0581-D1 (the field REALLY struggled on this one while Foldit takes #2, 4, 6, 7, and 8).

==> To me, this seems to indicate that folding by humans is the most useful when computerized methods fail. The question, then, is perhaps finding out WHEN computerized methods are likely to perform badly…

[Madde]

In the default ranking Foldit players take #1 and #5-8 on T0581-D1, you sorted them by AL0_P.