puzzle picture
1412: Sketchbook Puzzle: Unsolved Denovo Freestyle 105
Status: Closed

Summary

Name: 1412: Sketchbook Puzzle: Unsolved Denovo Freestyle 105
Status: Closed
Created: 08/02/2017
Points: 100
Expired: 08/09/2017 - 23:00
Difficulty: Intermediate
Description: Focus on your early game as we revisit puzzle 1381: Unsolved De-novo Freestyle 105. In this Sketchbook puzzle, you have 250 moves at your disposal. Once you use them up, you can reset and try something else! TELEKHREELKEFLKKEGITNVEIRIDNGRLEVRVEGGTERLKRFLEELRQKLEKKGYTVDIKIE
Categories: Overall, Prediction

Top Groups

RankGroupScorePoints
1Anthropic Dreams9,473100
2Gargleblasters9,43576
3Beta Folders9,42056
4Void Crushers9,29041
5Marvin's bunch9,28229

Top Evolvers

Top Soloists



Need this puzzle? Log in to download.  

Comments

gitwut's picture
User offline. Last seen 6 weeks 6 days ago. Offline
Joined: 05/18/2012
Groups: Contenders
Categorization of Sketch Puzzles

Frankly, I think the Sketch puzzle category should be lumped in with the Beginner puzzles or with the Tutorials. I don't think they should contribute to Global points as they lack any obvious scientific contribution. My opinion, of course, so I'm open to discussion to the contrary.

I suspect the puzzle type may have something to do with wanting to make Foldit available on touch devices? If that is the case, why not just say so to begin with?

jflat06's picture
User offline. Last seen 9 hours 56 min ago. Offline
Joined: 09/29/2010
Groups: Window Group
.

We are actually running this puzzle specifically for scientific purposes. From analysis we have done on previous puzzles, we are finding that most of the topological changes and valuable work is done in the earlier stages of folding. With that in mind, we're interested in a puzzle that refocuses folding effort on the early-game, rewarding you for trying something new instead of drilling on an existing fold.

gitwut's picture
User offline. Last seen 6 weeks 6 days ago. Offline
Joined: 05/18/2012
Groups: Contenders
Somewhat clearer

This would have been useful information in the earlier puzzle descriptions. My earlier understanding was that the emphasis was on economy of moves and maximizing score, not diversity of poses.

If you've found that the valuable work is done early, I'm guessing that it can't be extracted from previous puzzles? If not, why not change the way all puzzles are currently stored so that the data can be extracted from all forthcoming puzzles? That would seem more useful than creating a whole new category of puzzle.

Is each solution sent back for each Puzzle Reset, or is all the information stored and sent back when the puzzle closes? I ask because the client is extremely unstable when the scoreboard gets out of sync.

gitwut's picture
User offline. Last seen 6 weeks 6 days ago. Offline
Joined: 05/18/2012
Groups: Contenders
As I expected

I created one pose, no scripts, saved my work and then tried to open shared folders. Client locked up. These puzzles are entirely inefficient effort-wise and time-wise due to the far too limited move constraints and buggy client. I could have accomplished the same thing handfolding only in 1/4 of the time if I weren't so worried about minimizing my moves. The way I approach denovos (minus the move constraint), is already the same, but with better results.

I thought it only fair to at least try this puzzle type and, sad to say, most all of my concerns appear accurate. This puzzle type, and the client, needs serious re-consideration and re-working. Until then, I won't waste any more time on them.

jeff101's picture
User offline. Last seen 1 hour 37 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
Mining data to get multiple distinct solutions from each player:

My impression is that when a player plays online, Foldit regularly sends that player's latest (not necessarily best-scoring) solution to the server, so there is a history of moves for each player's solutions stored on the server. This should include for each player many solutions that are not their best-scoring solution, solutions that in fact look quite different from their best-scoring solution.

Despite this, I think when you analyze the results, you tend to only use the best solo and best evo for each player and the best solution from each group. If a player has explored a variety of structures, the lower-scoring ones are all ignored during analysis. This seems wasteful to me, especially with this new kind of puzzle. Do you plan to change your analysis procedure to make better use of all the data this new type of puzzle will create?

frood66's picture
User offline. Last seen 17 hours 16 min ago. Offline
Joined: 09/20/2011
Groups: Marvin's bunch
I think, with respect, u r

I think, with respect, u r missing an important point. Many hand folders work at low cI and Low wiggle power for a long time B4 they 'move up'. They do all this in 'blind belief' that they are creating a good fold. Doing this strategy takes a great deal of moves. So - with respect - I think u totally misunderstand.

gitwut's picture
User offline. Last seen 6 weeks 6 days ago. Offline
Joined: 05/18/2012
Groups: Contenders
More questions

It would also be more helpful to have the filter show something like "Moves made: 0, moves remaining: 250" or similar, since I imagine different puzzles might vary the number of maximum moves allowed. It would help to see that reflected by the filter.

I frequently run multiple poses on each denovo and monomer puzzle. I had four different track poses (much friendlier than reset, btw) for denovo 1405. Are all of them analyzed? Why can't the same data be extracted from the early folds of that?

Will all tracks of Sketchbook puzzles analyzed? If so, I'd prefer to use them instead of reset.

"With that in mind, we're interested in a puzzle that refocuses folding effort on the early-game, rewarding you for trying something new instead of drilling on an existing fold."

I'm not sure how allowing us to reset a puzzle and starting over with limited moves is a "reward". We've always had the ability to reset a puzzle.

Joined: 09/24/2012
Groups: Go Science
The number of players

For us, "early game" (hand fold and "observing" the protein) is the most (human) time consuming. This is also our most valuable contribution for this project. Overnight scripts or "moves" are almost "for free" (concerning Human resources). So the question is not the number of moves but to encourage us to spend more "hand fold" time.

Suggestion (just brain storming)

Game in 2 (3) rounds:
1-one round pure hand fold and no score (no rank, no points, no scripts unless GUI, no share)
2-one almost normal round with possibility to load solutions from round 1 in different identified unique tracks- At the end of this round, the score is not immediately calculated according to the apparent top solution's rank BUT ...
3-A definitive score and ranking "round" where the player's score and rank is the sum of the 2 current best separate original solutions (at an unique separate paths starting in round 1 ending round 2).

In round 1, my strategy would be to try as many and diverse manual designs as possible, regardless of scores.

In round 2, I'll develop the full potential of several of them in order to also find the second best one.

I use the "decreasing return" property of the puzzles. If I see that my best solution is near to its mojo (gaining few points a day), I'll be encouraged to concentrate on second best solutions with higher "points a day" potential, in order to maximize the sum of my 2 best separate solutions.

Not sure it works. Just thinking.

LociOiling's picture
User offline. Last seen 1 hour 34 min ago. Offline
Joined: 12/27/2012
Groups: Beta Folders
lots of quirks in counting moves

I'm finding that even informational recipes like print protein 2.5 and TvdL Show Worst 1.1.3 cost a couple of moves. (I'm looking into that.)

Also as previously noted, opening a cutpoint is one move, and closing it again is another.

Bands don't cost anything.

Wiggle costs one move, but you can wiggle for as many iterations as you like. And you can fiddle with the CI while you wiggle.

Moving a cut section costs a move. Every single little adjustment costs another move. It would be nice if placing the move tool on a section cost one move, but allowed you to move it until you're satisfied and clicked on another section.

jflat06's picture
User offline. Last seen 9 hours 56 min ago. Offline
Joined: 09/29/2010
Groups: Window Group
.

Thanks for the feedback!

We understand there are a lot of kinks to iron out with what is and isn't counted as a move, and we'll review any tools that players highlight as being potential issues.

For the scripts - we'll look into those.

For cuts - are you still seeing the cut insert and delete being 1 move each? We had released a fix for this in devprev, but perhaps it was undone.

For bands - we thought it appropriate that they cost nothing.

For wiggle - we are looking into changing this to 1 move = 1 iteration, but we want to make sure that the GUI has the capability of only running 1 iteration of wiggle before we can do this. The CI fiddling might be an issue, too.

For the move tool - that is a great idea. As long as you're just doing a string of rotations/moves, I can agree that you shouldn't be penalized repeatedly.

LociOiling's picture
User offline. Last seen 1 hour 34 min ago. Offline
Joined: 12/27/2012
Groups: Beta Folders
closing cuts is free...

jflat06 is correct, *closing* a cut no longer costs a move. Opening a cut is still counted as a move.

LociOiling's picture
User offline. Last seen 1 hour 34 min ago. Offline
Joined: 12/27/2012
Groups: Beta Folders
closing cuts is free...

jflat06 is correct, *closing* a cut no longer costs a move. Opening a cut is still counted as a move.

LociOiling's picture
User offline. Last seen 1 hour 34 min ago. Offline
Joined: 12/27/2012
Groups: Beta Folders
toggling filters costs you

Recipes like print protein 2.5 and TvdL Show Worst 1.1.3 use behavior.SetFiltersDisabled, which costs a move each time it's called.

Toggling filters manually does not have the same effect.

I've opened a feedback:

http://fold.it/portal/node/2003998

eusair's picture
User offline. Last seen 1 week 6 hours ago. Offline
Joined: 01/09/2015
Groups: None
.

"We are actually running this puzzle specifically for scientific purposes. From analysis we have done on previous puzzles, we are finding that most of the topological changes and valuable work is done in the earlier stages of folding. With that in mind, we're interested in a puzzle that refocuses folding effort on the early-game, rewarding you for trying something new instead of drilling on an existing fold."

if i am missing something here, please enlighten me. but atm, this all seems rather straightforward in my mind given the statement above...

i think you are going about this the wrong way; you're method is not going to produce the desired result. by allowing a limited number of moves, you are encouraging players to become more efficient with each move. every time they discover an improvement, if they feel so inclined, they can simply undo and find a more efficient way of achieving those previously discovered gains. the result is thus more time spent to go the same distance and the highest scores will be had by the most dedicated/efficient players, not the most skilled or visionary folders. additionally, you are unnecessarily opening a whole new set of problems by trying to define what is and is not a move.

there is a far simpler and better way to do this. if you are so very interested in the early stages of folding, then limit the puzzle to early game play only, such as with a time limit. give players ~2 hrs max to do as much as they can however they want to do it. the only thing that resets the timer is resetting the puzzle.

jeff101's picture
User offline. Last seen 1 hour 37 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
Finding the most efficient folding pathways:

Somehow, despite the many possible structures for a given protein sequence, nature quickly folds a protein into a specific stable structure. See https://en.wikipedia.org/wiki/Levinthal%27s_paradox for example.

In this sketchbook type of puzzle, I can see a player taking his/her best solution and trying to fold it again with a smaller number of moves. Then the player can use the extra moves remaining to find an even better solution. In effect, this will encourage players to explore the folding pathway and find the most direct ways to reach the best folds. The most efficient series of moves for the best solutions would make interesting movies that could help train human folders and might even resemble the true folding pathways used in nature. They would also make useful training sets for artificial intelligence and neural networks. See https://fold.it/portal/node/2002918 for example.

alcor29's picture
User offline. Last seen 7 hours 16 min ago. Offline
Joined: 11/16/2012
Thanks Jeff

Thank you for posting this Jeff. It makes me feel somewhat less ignorant. Though not remotely connected to biochemistry I have now long wondered if all the grinding done for points creates a false gamification artifact as opposed to a greater scientific validation. The path that is described by Levinthal coincides with my speculation that a string of amino acids may fold unto itself in accordance with the first likely encounters of the intermolecular forces. This has meant for me, that when we concentrate on grinding points for the game competition by ensuring that we are not in a "local" minima, we may actually be kicking the fold out of the local minima which the protein "prefers" in the real world of its milieu. I wonder how many other folders have felt the same thing.
( https://en.wikipedia.org/wiki/Levinthal%27s_paradox )

eusair's picture
User offline. Last seen 1 week 6 hours ago. Offline
Joined: 01/09/2015
Groups: None
.

all this folding pathways stuff does not require 'moves' in order to be studied. assuming pathways can be studied at all (see below) in foldit, it could just as easily, no, more easily be studied using a simple timer.

keep in mind the fact that nature doesn't have bands or cuts or idealize structure buttons; nor the hands and mind of a human player/chaperone. foldit is not a folding simulator. protein structures are not arrived at in a way that at all resembles that in nature.

frood66's picture
User offline. Last seen 17 hours 16 min ago. Offline
Joined: 09/20/2011
Groups: Marvin's bunch
Since (it would appear) we

Since (it would appear) we are going to be spending more time working out how many 'moves' we have made (some of those that count appearing apocryphal at best) Would it be unfair to ask the science behind this latest puzzle type. So far as I can see at the moment, it looks like a game for game's sake.Or as many B4 have described similar ideas....another 'Lab Rat Job'

Sorry if that sounds harsh - but it's only history.

eusair's picture
User offline. Last seen 1 week 6 hours ago. Offline
Joined: 01/09/2015
Groups: None
.

aye, its merely a game within a game. let's find out who wants to spend the most time getting the most out of each and every move. atm that means before you shake, add bands to every aa you want to end up a particular way. same thing with every wiggle. believe me, i've tried it. i've spent 20-30 mins preparing for a single shake. i press the Z button enough as it is. i don't see myself carrying on in that fashion for no reason, or for what seems to me atm to be unfounded and ill-conceived reasons.

Joined: 06/24/2008
Groups: Void Crushers
Blue Print

I think it would be good to have Blue Print for these types of puzzles.

Aubade01's picture
User offline. Last seen 8 weeks 1 day ago. Offline
Joined: 10/12/2013
Groups: None
It appears obvious that all

It appears obvious that all players of this Sketchbook puzzle are not on an equal footing. Players should start equal since moves are limited. Since this puzzle is based on the old puzzle 1381: Unsolved De-novo Freestyle 105, players who have the top fold pose stored on their computers, or are members of a group who have that top pose, have the advantage. Players who did not discover the top pose in 1381 or are trying it for the first time cannot possible score well in 250 moves. If Foldit is really serious about Sketchbooks starting as old De-novo, then a link to the top pose image should be available. We should all know at the start how the coils and sheets are aligned.

>>>>> Aubade00 / 01

Susume's picture
User offline. Last seen 2 hours 10 min ago. Offline
Joined: 10/02/2011
See blog for equal footing

There is a picture of the protein in the blog for all to see - scroll down to the one from Waya, Galaxie, Susume from puzzle 1297.

Aubade01's picture
User offline. Last seen 8 weeks 1 day ago. Offline
Joined: 10/12/2013
Groups: None
Puzzle appears to be a design

Puzzle appears to be a design not De-novo.

"1297: 65 Residue Monomer Design: No Rebuilding!"

Blog picture 1297 has only a single coil not the two in Sketchbook.

I was thinking more along the lines of a picture link embedded in the puzzle description. We should be encouraging new players and not expect them to know everything veterans know.

>>>>> Aubade00 / 01

LociOiling's picture
User offline. Last seen 1 hour 34 min ago. Offline
Joined: 12/27/2012
Groups: Beta Folders
the first successful Foldit design...

The protein in this puzzle is the first Foldit player design to be grown, crystallized, and solved by X-ray diffraction.

As Susume mentions, the protein was created in puzzle 1297. After the wet lab, it was presented as a de-novo in puzzle 1381, and an electron density in puzzle 1384.

See Design Puzzle Results on the wiki, which includes a nice image posted by Waya for 1297. The Puzzle 1381 gallery shows the protein from two angles.

I'm still not convinced about the sketchbook format, but we've had lots of practice with this particular protein already. I'm sure it will end up the PDB someday soon.

Get Started: Download
  Windows    OSX    Linux  
Windows
(Vista/7/8)
OSX
(10.7 or later)
Linux
(64-bit)

Are you new to Foldit? Click here.

Are you a student? Click here.

Are you an educator? Click here.
Search
Only search fold.it
Recommend Foldit
User login
Soloists
Evolvers
Groups
Topics
Top New Users
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, Microsoft, Adobe, RosettaCommons