Idea: a new mode, for the good of science

Case number:699969-2004810
Opened by:Vincenzo Brancaccio
Opened on:Wednesday, February 14, 2018 - 00:17
Last modified:Thursday, February 15, 2018 - 20:44

What a folder does is finding the most stable conformation of a protein, by avoiding a forest of "local energy minimums", armed with algorithms, experience, and possibly a bit of luck. Let's call it a sort of "brain-assisted brute force" approach.
We know that there are already some teams which, as far as I know, can share protein structures and travel together towards the "global energy minimum" from the best minimum they could find, which might or might not be an advantage: maybe further exploration from that minimum could be uneffective, or it could lead to better structures.

I will split the idea in two parts: the "scientific" part and the "gaming" part:

1. The scientific part
As the times left for a puzzle narrows, new users are always given the same starting structure. The idea is perfect if want to give easily an equal score to the users, but it is not very efficient, if you consider that the chances to get a good structure are poor for people beginning the puzzle some time later.
Picture yourself a tree graph. The highest dot is the highest structural energy (the usual beginning structure). The energy decreases going down, and any point is an energetic minimum (best structure form the users). Any point is connected to the dot it comes from: if you ignore "single-user evolutions", any point is connected to the structure it comes from. Any structure derived from the first structure is directly connected to it, any structure derived from a shared structure is directly connected to it.
As the time proceeds, new users could be given the choice of starting from the "usual" basic structure, or to be assigned a random structure so that:
-the assigned structure IS a parent (or parent of a parent) structure of one of the current best structures
-the set of assigned structures is NOT son of the same parent structure
-the assigned structures are energetically similar (energy of choice decreases as the global game and structures proceed)

2.The gaming part
Since players would probably like to be given the deserved contribution, some points could be assigned by considering the generational contributions of any user who contributed to the formation of a "branch" in the tree graph. If user A gets in one day the highest score by using a structure obtained by user B in three days, with a score progress of 20% compared to user B, then he is assigned 20% of the points, while user B is assigned the 80%.
Also, a user "stuck" in a bad structure, might be given the one-time option to randomly switch to one of the random assigned structures. Or maybe the option could be re-activated anytime the user reaches some target score.

What advantages do I see in the idea?
1-Better employment of "computational time": any user, joining the game anytime, could give important contributions towards the best structure, which almost certainly couldn't be possible with the same beginning structure for anyone
2-Possibility to discard bad structures, and to focus on "new bests" for anyone, even if not in a team.
3-Being assigned structures from different families, exploration is directed towards different minimums. This already happens when the beginning structure is the same for everybody, but the proposed system might be much more efficient.

(Wed, 02/14/2018 - 00:17  |  4 comments)

Joined: 09/24/2012
Groups: Go Science

This is a refinement of the idea discussed here:
(up to #20 - The end of the discussion is about a different subject).

It's interesting to read the discussion there before to answer here.


One option is of course to join a group: within a group you can start over with one of the best group designs ... but you are limited by the 2-3 promising group designs available.

With your idea, any player could start from a larger variety of promising good designs.

The only problem is to "keep" the secrets of the "good ideas" until the end of the puzzle, otherwise there is a risk that everybody stick to a limited number of initial "false-good" designs. For example, there is a risk that a member of a group, by resetting, discovers the alternative design from another group or from a soloist. This could make converge the end solutions to a specific non-optimal "best".

Hanto's picture
User offline. Last seen 10 hours 15 min ago. Offline
Joined: 05/10/2008
Groups: None

Might consider joining Go Science ( not my group ) as your comment indicates an open mind sorely lacking in current crop of team leaders and PwP (People with Power ). Watch for sensitive toes and thin skin whatever you do. I would have marked your comment up but can no longer see the arrows well enough. Use your open mind well.

Joined: 06/04/2015
Groups: Go Science

Sorry for the late reply, thank you both for your answers, and sorry if I brought up such an old "problem"!
I will surely look up the team you suggest and try to understand how teams work!

S0ckrates's picture
User offline. Last seen 8 hours 35 min ago. Offline
Joined: 05/19/2017
Groups: None

It does sound more efficient but the inherent trade off is the loss of branches players can take in the folding stability funnel over time. Personally I think because of the *competitive aspect* players should always start on equal footing regardless of when they start within the puzzle time period, so that the competition is not affected by a player who starts with a partially-pre-folded structure with an inherent advantage (or disadvantage, it goes both ways.) Now, of course, this game isn't pure competition, it's a sort of hybrid in between, so as it stands I think groups are kind of the compromise for this. I suppose therein lies the challenge: I hope there's a group out there right for you!


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, Microsoft, Adobe, RosettaCommons