17 replies [Last post]
phi16's picture
User offline. Last seen 2 hours 24 min ago. Offline
Joined: 12/23/2008

Self-fulfilling prophecy = Fold.it player strategies take their current form only as a reaction to the tools given.

Could it be that what we are seeing as a player strategy is nothing more than a reaction to the tools that were provided?

If the tool set given by Fold.it was different, the game playing/protein folding strategies would be very different. So what do we hope to learn here? How important is it to develop a broader set of folding tools with many different strategies and techniques.

For example, the game only saves a top score. If you make a bad move (loose many points by stressing the backbone in various ways) as you struggle back, resolving the issues in neighboring pairs, you might make some very good, new folds. However, if the overall score is not a new high, any interim work is lost should you quit before reaching a new high. So.... the only worthwhile increase producing scores are the small incremental scores you might see by walking the backbone (wiggling two or three segments at a time ONLY). Likewise, any serious large scale rebuilding is totally impossible unless you are successful. Not many will be fortunate enough to achieve a new top score after a major rebuild of more than a segment or two at a time.

Imagine if there were many different folding tools? I could name a dozen.

Joined: 05/30/2008
Groups: Contenders
Then name them

The feedback tracker offers ample opportunity to submit suggestions for new tools. As a matter of fact new best has already been implemented in the developer's preview from what I understand.

axcho's picture
User offline. Last seen 7 years 27 weeks ago. Offline
Joined: 12/18/2007
Groups: None
Interesting supposition

I wouldn't be surprised if different strategies would be dominant if the tools were significantly different. Can you think of an example scenario where a new tool or scoring system resulted in a certain new technique? I'm curious to hear what you have in mind.

Joined: 05/19/2008
Groups: None
I'd recommend saving games

I'd recommend saving games more often, phi. You can enter your notes in the save dialog, so you always can find your way back. That's what I do at least.

Axcho, regarding a tool that results in a new folding strategy: there are plenty of suggested tools in the feedback section that would do that. For example the ability to fold the backbone as a wire, while calculations are turned off. This particular tool would significantly reduce the time to achieve the structure that seems perfect for a molecule.

axcho's picture
User offline. Last seen 7 years 27 weeks ago. Offline
Joined: 12/18/2007
Groups: None
I know about the tools

I know about the tools in the feedback tracker. I was more interested in hearing phi16's ideas on how the strategies would change as a result of different tools.

xiando's picture
User offline. Last seen 4 years 40 weeks ago. Offline
Joined: 10/19/2008
Groups: Oma Gawd
It makes sense

it makes sense that using the tools shapes the players methods, much as an artist's palette of colors and his materials shapes his work. (Or hers...as the case may be)

there are limitations presented by any toolset, as there are opportunities.

regarding states: you are aware of the intermediate save states provided by CTRL-SHIFT 1/2/3 and their respective restore-state functions CTRL-1/2/3 yes? this may go some way to providing you with further abilities...as does the recent inclusion of an intermediate "best" value.

Agreed on the dozen comment...I think I have that many in the que right now... ;)

phi16's picture
User offline. Last seen 2 hours 24 min ago. Offline
Joined: 12/23/2008
Then name them -- I'll be happy to name a few.

I'll be happy to name a few new tools but I want to make sure you understand the problem I'm pointing to. If people are always going to return to best score then it is very difficult, if not impossible to start a rebuild which involves going back down in score perhaps 1000's of points in order to arrive at a new best. As soon as you give up on a local attempt all of your interim work is lost when you jump back to best score. So image going down 1000 points, finding a better solution which gives you a new 500 points in a different area, and then having to give back that 500 points when you hit last best button?

No wonder the strategies people gravitate towards and use are the ones that yields little bitsy gains at a time like walking the backbone?

phi16's picture
User offline. Last seen 2 hours 24 min ago. Offline
Joined: 12/23/2008
Interim saves.

You make a good suggestion but I'm afraid it wouldn't help.

Imagine going down 1000 points when you try a major rebuilding of a section?

You go down 1000 points.

You rebuild.

You go up 700 points because of that exciting new work you did.

But it is not a new high. :( All is lost even if you save. When you hit best score button. All your good work goes away and you have to start again.

phi16's picture
User offline. Last seen 2 hours 24 min ago. Offline
Joined: 12/23/2008
Suggested new strategies

I'll give a stab at suggesting new strategies. Your idea about using a wire structure while turning off calculations seems like something in the right direction. But I don't see how it could be very useful if it doesn't do any calculations. You know how these things just fly apart when something doesn't fit well.

xiando's picture
User offline. Last seen 4 years 40 weeks ago. Offline
Joined: 10/19/2008
Groups: Oma Gawd
Hmm...

I'm not sure you understand the interim "best" and disagree that it results in the problem you describe. in fact it was envisioned to assist in avoiding that situation. And AFAIK, from other feedback I've received, it is already being used to that purpose. The only major compalint I've heard is that tsome players don't like the hotkeys used.

Further...If you choose to do something that forces the puzzle score to a low level, you have likely changed the entire protein, although you may not think so. The lost points are often not just in the area you work, but also in residues at the farthest dimensional point from that area. So a cut and paste approach (am I seeing what you're talking about?) isn't going to work. Free energy (score reduction) is not injected into the protein at only the point worked, but often in each and every residue...check and you'll see... incremental scores for many or all of the residues can be negatively impacted by major rebuilds.not just the localized structures thru the process of equilibrium (equilibration?) during global wiggles subsequent to the "Rebuild".

Intermediate saves using recent best should maintain the scoring so that hitting restore recent best calls the value,  UNLESS you make the mistake of restoring the global or "Very best" solution, which by the fundamental nature of the compare/replace functions used for both VERY best and recent best require that replacement, or if the value is exceeded by work subsequent to setting the baseline point. It was never meant to act as CTRL-SHIFT 1/2/3

If you choose to work an area in which you want to save the low score states regardless of their value wrt the the highscore, ctrl-shift 1/2/3 were designed for the very purpose.

if you want to work from a ctrl 1/2/3 state, then set the recent best marker from that location...ie, recall the location using ctrl 1, 2,  or 3 and set recent best. All subsequent calls using restore recent best will be to that value or the value automatically replacing its initial value when it is exceeded by your work, whether that is by incremental score increases or by recalling the VERY BEST score.

And to the best of my knowledge, the recent best value isn't deleted during a file save operation any more than it is during a file load. If so, that is a problem/bug.

It is only when a score exceeds the recent best set point that it's value is overwritten.

Recent best is like an allen key screwdriver...not a multitool. As are
most. Like programs...No one program can do it all.  The wiki has a
description, albeit primitive, of how to use the recent best tool.
please read it if you have time. It somewhat spells out what it does by
example. If you sitll find it confusing, I'll try to revamp the entry
to be more palatable.


I guess am unequipped to address the fundamental question. Only to agree with it's basic premise that methods are influenced by toolsets...it's a no-brainer...

And of course attempt to help clarify tools that may have been mis-interpretted by players who are simply using a new feature for which they have no historical background (the topic of a secondary restore best point has been ongoing for over six months since I first proposed it, but you were not hereto benefit from those previous explanations)  Perhaps someone else can better explain it to you if I have been unsuccessful. It has limits but it has strengths too.


xiando's picture
User offline. Last seen 4 years 40 weeks ago. Offline
Joined: 10/19/2008
Groups: Oma Gawd
Axcho: I can

I can think of a concrete example to answer your query above.

I once proposed the inclusion of the ability to clear bands during the wiggle using a hotkey for the express purposes of allowing snap functionality to the folding process early on in the public release. it was accepted and implemented.

The reason that the R bug was so prominent? Because a whole lot of people rely on this ability for folding.

it's a silly "tool" but it IS a tool. and imo although perhaps an insubstantial tool when compared with a more obvious choice like "Rebuid" or Rotate", a valid example of how a tool changes the way people work. I believe it also meets the implied requirement in your query of being a tool that actually exists and who's use can be measured and proven, rather than a hypothetical situation, as would be the case for any proposed but un-implemented suggestions located in the feedback tracker suggestion bins, as their beneift to the community is unsubstantiated except by hypothesis, promotion,  and a boat load of background work to back the first two up.

regards,

phi16's picture
User offline. Last seen 2 hours 24 min ago. Offline
Joined: 12/23/2008
New Folding Tools Suggestions -- 1. Change the scoring.

To continue the dialogue by making some suggestions to get around the problem I point to in this thread I would like to say:

1) I avoid reading other people's ideas until I have given myself a chance to think through the problem. I do this by training. I have found I am too likely to get locked into other's thinking if I look at their solutions before I come up with my own.

2) I don't know much about proteins or protein folding so I might have to ask more questions than provide answers while thinking this through.

That having been said, please forgive me if my contribution is redundant or just not workable due to the nature of protein folding that I am unaware of. I would love to do more reading on the subject and welcome any suggested reading.

OK. My first suggestion is simply to change the scoring. It might not be as pleasing or fun to play but if the scoring was done differently, it would encourage people to solve differently.

For example, add the rule --"Increases in scores can only be achieved in 100 point increments." Small incremental advances wouldn't count until a player could put together 100 points worth. In this case, people would opt for the major rebuilds and spend several hours trying out different strategies to achieve them rather than wiggle pairs of amino acids. In so doing, people will focus on the types of tools that they would need to better do this sort of folding, e.g. better FREEZE, LOCK and BANDing. (See 2. FREEZE, LOCK improvements)

phi16's picture
User offline. Last seen 2 hours 24 min ago. Offline
Joined: 12/23/2008
New Folding Tools Suggestions - 2. FOLD, LOCK, BAND Improvements

2. FOLD, LOCK, BANDing improvements

In order to do major folding, taking large chucks of protein strands and rearrange them, the user needs the ability to truly lock together a section such as a helix, and treat it as a single entity to be able to place it in a new location, all at once. To do this freeze or lock should be available to lock a section together, BUT allow the section to be moved around in space. AND, bands should allow a user to guide such a large section into a new position.

To improve the ability of folders to rebuild in more profound ways than just a few segments at a time, the tool set should include better ways of handling larger sections. The problems with existing tools are:

1. There seems to be some confusion of the functions of these tools. As is, they don't match the intuitive understanding a player would assign to each. FREEZE, intuitively, should freeze the amino acid in every way. It should freeze its position in space. It should freeze its relationship to its immediate neighbors. LOCK, intuitively, is a different name for the same function and should do the same. Sometimes they perform as expected, sometimes they don't.

2. Banding doesn't work well. It's confusing. Is it allowing the user to tie the amino acid to a fixed point in space, OR, tie it to a fixed point in relative relationship to the protein, e.g. a fixed distance from the backbone. I'm not sure what Banding is supposed to be doing. I understand when tying together two amino acids with a band and there is a pull between them that has been created. But I don't understand what's going on when banding an amino acid to a point in space. Sometimes the pull is pulling the acid to a point in space. Sometimes that point in space seems to remain fixed, causing the entire protein to twist and turn. To combat that effect, I have tried to put the amino acid between two equal bands. I'm guessing, that because I did not setup that pair of bands exactly equal, the amino acid is pulled toward one and away from the other. But in either case, the entire protein is pulled out of wack with little control by the user. I also don't understand why banding cannot be used to pull on two neighbors.

SOLUTION

I would like to see FREEZE divided into two functions: freeze the relationship of an amino acid with its local neighbor, and, freeze the amino acid in space. The user should be able to do both or each function separately. I would suggest using two words to equal two functions: FREEZE = fixed in space. LOCK = fixes relationship with neighbors. In this way a section could be LOCKED and moved around as a block.

I would like to see BANDing improved by also being divided into two functions. Banding should allow user to tie an animo acid to a point in space. OR, banding should allow user to tie an amino acid to a fixed point in relation to the backbone. If we give the backbone a direction, e.g. start to finish. And give backbone an angle reference 0-360 degrees a band could be described with two angles, e.g. 0-360 around backbone and +90 to -90 from start to finish. This would allow banding to be used to straighten a group, for example.

phi16's picture
User offline. Last seen 2 hours 24 min ago. Offline
Joined: 12/23/2008
Hmmm......

Hi Xiando. Thanks for your comments. I believe I do understand INTERIM BEST. It will prove to be a useful tool, albeit a little more difficult to use than VERY BEST. It is a step in the right direction, however, it won't help solve the problem I'm referring to. Here's why:

Let's say to do a major rebuild of a section and loose 1000 points before you start shaking out the new tensions created during the rebuild. You do a good job and recover 700 points THAT YOU PREVIOUSLY DIDN'T HAVE. By that I mean those were just not recovered points but new found points!.

Unless you are able to move the overall score over the old high score you will not be allowed to keep these points unless you never use VERY BEST again. And you and I both know that using VERY BEST is a oft used function we could hardly do without.

xiando's picture
User offline. Last seen 4 years 40 weeks ago. Offline
Joined: 10/19/2008
Groups: Oma Gawd
a couple more are in order

1. Suggestions need to be sent to the feedback tracker, not posted here. The development team does not really pay attention to this forum regarding suggestions to improvement, as it is not the forum in which that type of discourse is meant to be posted, and the content cannot be assigned or tracked when players submit ideas in the wrong location.

2. re: Very best and recent best.

I don't think you quite get it yet. You have no idea that you "found new points", you just found 700 points. Prior to that you reduced the stability of the protein and lost 1000 points, therefore you have simply regained 700 of the 1000. These are not a provably "new" source, as you have not exceeded your losses and you would have a score 700 points above your original.

back to the perplexing issue...

Very Best calls back the very highest score. nothing else. That is it's purpose...to track and store the very best score. it values is updated by a comparison with the current score, and it is otherwise static, ie, it cannot be reset by the user without playing tricks that were unintended by the developers

Recent best call back the best score achieved following application of a user-specific setpoint. That's it's purpose...to track and store the best score achieved following a reset (Set recent best) event.

Both behave identically from a mathematical standpoint, except that Recent Best tracks the score from a user-specified starting point by setting that start point using the "set recent best" command.

I simply don't see that the situation you're describing is overlooked here. In fact, that situation is WHY recent best was originated. Please read the wiki entry on this feature, as the situation you was specifically addressed in that entry.

I'm not sure why you're not getting it, but I'd suggest that you talk to others in the chat to resolve any further confusion on your part as to what is possible and what isn't using very best and recent best. Perhaps I'm missing some fundamental point you're making, but I really don't see the issue.

Joined: 05/30/2008
Groups: Contenders
100 point gains

This actually sounds like a good idea to me, so long as it was added as a separate stage in the folding. Have both, first release the protein requiring 100 point gains then release it with standard parameters. This would of course be accompanied by a separate ranking system.

I suggest you post it in the feedback tracker, or if you don't want to do you mind if I do?

phi16's picture
User offline. Last seen 2 hours 24 min ago. Offline
Joined: 12/23/2008
Change in Scoring

Ty ZZZ. Xiando has been encouraging me to move these suggestions over to the feedback tracker so I will.

phi16's picture
User offline. Last seen 2 hours 24 min ago. Offline
Joined: 12/23/2008
Using Bands to Compress the Protein

Using Bands to Compress the Protein

A good technique, early in a protein puzzle is using bands to compress the protein, making it more compact. To do this, randomly click an amino acid segment while holding the shift key and drag the newly formed band to another segment. Let go when the cursor is over another segment and a band is formed between the two. Try three or four bands stretching from one side of the protein to the other. Next, hit "W" to do a global wiggle. Experiment with how long you allow this to continue before stopping the global wiggle by hitting SPACE BAR. Hit "R" to remove bands and hit "W" to global wiggle again. Watch scoring. If it slows and stops without hitting a new high, hit "A" to restore best. If it reaches new high, VOILA! You've successfully used bands to compress the protein. You might want to try it again.

This script times out to 60 seconds.

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, RosettaCommons