11 replies [Last post]
Joined: 03/05/2015
Groups: Gargleblasters

Foldit scoring, when it uses filters, can be discouraging, frustrating, and downright annoying (not to forget slow, sometimes VERY slow).

The latest filter, Ideal Loops, was the "last straw" ... and on puzzle 1269, when combined with the Sheets filter (and how the IL filter sees sheets that are missing bonds as loops which is maddening), and the residue fiters, SS and structure filters...made it incredibly frustrating to find a good score.

Using game rationale, which I understand was to use positive points to reflect a lower energy score from a protein for a good fold... What if, Instead of showing up as penalties, the bonuses are there only as a reward for an even better fold.

I think that this is the intent of filters, but the way that they are currently scored makes them punishment until a folder does well. (stick followed by carrot, if you will).

I think changing this (to reward instead of penalty followed by reward) would encourage rather than discourage. I've scrapped so many designs for puzzles because they didn't meet the filters and that got me to thinking that there has to be a more positive way to get results.

What do you think?

Joined: 03/05/2015
Groups: Gargleblasters
Related: ONE (perhaps the ONLY) good use of the IL filter :0)

A little mandala art from MVY: Admire!

jeff101's picture
User offline. Last seen 18 hours 36 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
I like the idea of bonuses instead of penalties.

I don't usually do Design Puzzles or ones with multiple Filters, but
I do think having bonuses instead of penalties to encourage certain
designs would help.

I also like the Mandala design posted above.

Joined: 10/23/2014
Groups: Contenders
Making us loopy

I thought the purpose of ideal loops is to make a design that will actually fold in real life, that is in the wet lab. If the design has loops that aren't ideal the likelihood of it folding are diminished. Many top designs that went to the wet lab failed to fold hence ideal loops was born.

Everyone likes to get good scores but in the end it's the research that matters. My group has had talks and demos on how to create ideal loops and we have improved. Unfortunately it can still be laborious. Perhaps someone can do some videos and papers on the various techniques available.

Actually I would like to see better tools that would make it easier to create ideal loops. I have ideas and I'm sure other do as well on how to improve existing tools or create new tools. It shouldn't be this hard. Computers should be able to help simplify the process for us loopy humans.

Joined: 03/05/2015
Groups: Gargleblasters
You have some great suggestions Tom

And I definitely agree about the need for improved tools. Another thing that would help would be better tool tips (some are ambiguous and more confusing than helpful)

I do understand your comments regarding this being about science, and while true, serious work on scientific problems isn't mutually exclusive from making the game play more enjoyable.

I also love your idea to have more sharing of tips and folding techniques through videos and better tutorials and Susume is a great role model for that - she and other great folders have shared with us. That is one thing that the groups do well particularly within some of the groups that would be fantastic to make more available for the entire foldit community.

Joined: 03/05/2015
Groups: Gargleblasters
Examples of confusing tool tips/indicators

In the Conditions dropdown, for example... if you have too few sheets, and click to show them, the sheets that ARE counting show up with red blobs next to them.

By contrast, when you have non-ideal loops, and click to show them, the red blobs are next to the nonideal loops. Since red means "bad" in one case and "good" in the other, this is confusing.

If this were consistent, the good sheets would not be highlighted, only those segments you have designated as sheets that don't meet the filter.

If the blob for good sheets were a different color than red, or the tool tip said "show good sheets" this would be helpful.

The colors for sidechains in the residues are also confusing, for a similar reason. Good sidechains are brightly colored, but that's not quickly understandable from the conditions menu wording.

Little tweaks like that could make those filters much more meaningful.

Joined: 01/01/2016
Groups: Gargleblasters
Lots of good suggestions, of

Lots of good suggestions, of course we want better science, but it's essential that the gameplay isn't degraded to the point of not being a game, after all that's why a lot of players started with Foldit and we want to keep those players and encourage more.

My perfect solution for easier ideal loop formation would be something along the lines of - Identify an ideal loop, (either from a fold in progress or from a library) - copy it - choose where you want to put the loop (either highlight the segments or select and delete them) - paste the previously copied loop - voila!

Of course I've no idea how easy it would be to do that in practice, but there has to be a better way than the current situation.

Joined: 06/06/2013
Groups: Gargleblasters
I Deal Loops

since trying to work these properly has been so slow on my machine, I ended up picking a slow time and would run the remix tool until the filter got past. often it would be over 24 hours even for basic folds like laid out in Nature Article about ideal proteins. problem is then that can't let drw clean anything up as it will often mutate the loop to something that isn't ideal
I would like to see a library with a list of ideal loops. Not stick mode -- something we can all read and understand. then we could produce the folds wanted.
Until then, time for me to follow Kenny Rogers' advice and know when to fold em. I don't have enough CPU for these puzzles

Joined: 09/24/2012
Groups: Go Science
Ideal loops, rama, remix and wiggle

For ideal loops, I almost never succeed to make the short loops (2-3) between sheets ideal :P

a) I wonder why most remix proposals are not ideal.

b) Remixes are not always good on rama. Why ?

c) When wiggling a good rama section, most of the time, it makes the rama worst than before. :P

d) When mutating the loop to anything flexible (ala, gly etc), it doesnt' seem to work better :P

e) When reducing the CI, I didn't succeed yet.

f) When rebuilding a lot of time, I don't succeed to find ideal loops.

My only way so far to idealize loops is a lot of luck using recipes on loops (quaking remix) or any other rebuilding recipe (edrw etc).

But when looking on the results, I see that most top players got the 810 filter points. Could any of you give some tips to make the loops ideal ?

Thanks !

Susume's picture
User offline. Last seen 2 days 3 hours ago. Offline
Joined: 10/02/2011
(post moved)

(Moved my post on how to make ideal loops to its own thread, so as not to distract from the original topic of this thread.) http://fold.it/portal/node/2002700

bkoep's picture
User offline. Last seen 2 hours 2 min ago. Offline
Joined: 11/15/2012
Groups: Foldit Staff
Helpful suggestions!

I'll chime in with a few thoughts from the "scientific" perspective:

Re: bonuses vs. penalties
I absolutely understand the preference for filters that effect bonuses instead of penalties (i.e. carrots instead of sticks), and we could probably adjust some of our "discouraging" filters to work on this principle with minimal disruptions. In many cases, unfortunately, the "penalty model" better reflects the intention of the filter, and restructuring the filter would make it (even more) incomprehensible.

With the Ideal Loops filter, for example, we don't want players to design additional loops into their structures—so we can't simply reward each ideal loop. And one troublesome loop doesn't necessarily doom a good design—so we also don't want an all-or-nothing bonus that only kicks in at 100% satisfaction. The penalty model skirts both of these issues, and behaves in a way that is easy to understand (even if it is discouraging to lose points).

(For what it's worth, we expected the Ideal Loop filter to be extremely restrictive, and we've been very impressed at how many Foldit players have been able to design proteins with 100% ideal loops.)

bkoep's picture
User offline. Last seen 2 hours 2 min ago. Offline
Joined: 11/15/2012
Groups: Foldit Staff
Re: better tools

The Remix tool seems to help some of the problems we normally see in Foldit designs. We think that part of that is just because players are less frequently using Rebuild (which can introduce problematic backbones). If we can completely phase out Rebuild, we might be able to scale back some of the more stringent filters.

We'd really like to have a "copy/paste" function for ideal loops. We have some ideas floating around about how best to implement this, but it will probably be some time before we can get to it.

The feedback on tooltips is welcomed! Specific suggestions are especially helpful (like actiasluna's comment about the Secondary Structure Filter—this has been brought up before, and we should have it fixed soon).


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, Boehringer Ingelheim, RosettaCommons