No replies
Joined: 01/25/2008
Groups: None

Soon we will be seeing CASP puzzles that are not just for practice. These are proteins that no human on the planet knows the structures of. These will mark the first time non-professionals such as my self could ever have submitted a prediction to CASP. And to do it via a game that is fun and interesting to play is a great idea.

Here's the CASP8 website:
click the "target list" at top center to see the proteins being released for study. The idea is to challenge the scientific teams to use their current best methods of creating predictions and submit their best models prior to the completion of the native structure analysis done with the current scientific methods available.

These are x-ray crystallography, and magnetic resonance imaging. Very expensive and time consuming approaches. Especially when you consider how there are 100s of thousands of proteins in nature we would like to analyze, and literally trillions of billions more that are theoretically possible and might cure a disease or speed a chemical reaction that would be useful.

Here's an article that explained a lot about x-ray crystallography for your reference:

There has been much discussion about team sharing of puzzles, and with CASP we're all on the "" team and should work together to product the best model possible.

There are many approaches to handling the model sharing between players and trying to somehow reflect the effort produced in the player's global rankings. But I would assert that waiting to release CASP targets on the game, until these scoring issues are addressed is undesirable. It will take several runs using any new approach to global scoring to resolve the issues that arise. And you won't see the effects of any new approaches until the puzzles close.

In the interest of time, and encouraging full collaboration on CASP targets, I suggest that CASP target puzzles either never close, or otherwise that they be omitted from the global rankings, or seperated similar to how tutorials are seperated from competition puzzles.


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, RosettaCommons