17 replies [Last post]
Susume's picture
User offline. Last seen 2 days 19 hours ago. Offline
Joined: 10/02/2011

I thought people who are succeeding at making ideal loops might share their strategies here.

Rama instead of cutpoints
I have never tried making ideal loops appear on a protein that has already been built and wiggled - I make the ideal loops at the beginning by using the rama map to bend the straight protein into shape (instead of using cutpoints). I set the loop segments to glycine, aparagine or aspartate, then get the sheets and helices pretty close to each other by using the rama map on the loops (and often on one sheet or helix segment next to the loop). Then I freeze any long straight portion that I haven't folded yet, freeze the sheets and helices, and use bands to pull the structures into line. Once all the structures have been placed I move on to unfrozen wiggles, mutate, additional bands to increase the core, and scripts.

If you have an ideal loop and want it to stay, be sure to set the AAs before wiggling to something where each rama dot is in a good place for that AA - if a rama dot is in the green, set that AA to glycine, or to asparagine or aspartate if the dot is in the upper left portion of the green area. If you leave it as isoleucine or something else without a good green area on the map, wiggle will try hard to pull it out of the green shape.

TomTaylor tells me that he and others on his team don't use the rama map but are still able to make ideal loops, so there are multiple paths to success. Be sure to follow the "ideal protein design" rules about which way to turn your sheets and helices if you want to be able to add ideal loops later. If your sheets are turned the wrong way, or your helix is on the wrong side of the sheet, an ideal loop will not be able to form.

article on ideal design rules (only need to read figures 1 and 2 and their captions): http://www.nature.com/nature/journal/v491/n7423/full/nature11600.html
video on ideal design rules: https://www.youtube.com/watch?v=uXrQ2VWsPJ0

Joined: 09/24/2012
Groups: Go Science
Thanks again Susume !

So I summarize for myself:

1) Follow the "ideal turn" rules of the paper (even if it's a hard brain exercise for me)

2) Ideal loop from very start (few or no chance to correct it afterwards)
Several techniques are possible (rama map, freezes and bands, perhaps also cuts)
Using remix and take the one that rise the filter. Then only shake, no backbone wiggle neither rebuild.

3) Before to wiggle, mutate to the right glycine, asparagine or aspartate depending on the abeda colour on the rama map.

4) I suppose that it's better to have many points before to run recipes that will wiggle a lot. Or freeze the ideal loops at early game. I suppose that rebuild (no wiggle) is ok (rebuild is a kind of 3 segments remix as I understood).


In the past, I didn't follow your advice, paper and video on ideal loops, because I got good scores without applying this. But nothing useful for science as I now understand :(

Perhaps Foldit should add a loop filter for de novo too ?

Joined: 10/23/2014
Groups: Contenders
Quick and Easy

This is a quick and easy way for ideal loops but doesn't always work on every loop. Try another approach on any remaining non-ideal loops. Just tried this on puzzle 1272 with 6 loops and got all ideal loops in less than 10 minutes.

First Remix doesn't work on 2 or 3 residue loops. I use a minimum of 4 residues on loops. Set up the design as you would like it to look. Connect all cuts, freeze all sheets and helixes leaving loops unfrozen. Change all the loop residues to glycine, aparagine or aspartate (per Susume). In Pull Mode select Remix and hopefully there will be an ideal loop in the list. Note: sometimes the list has 0 choices. Continue through the loops. If you have any non-ideal loops left try the Rebuild option.

As stated earlier this doesn't always work but hopefully it will help.

Joined: 10/23/2014
Groups: Contenders
Addendum - Quick and Easy

It doesn't appear to matter which one of the three AAs you select, unlike what appears to be the case with the Rama Map. I haven't done extensive testing but if anyone finds it does please leave a comment.

Joined: 10/23/2014
Groups: Contenders
Update - Rebuild tool

The Rebuild tool is suspected of causing isolated cases of problematic backbones. It's now recommended Rebuild not be used to create ideal loops. Remix is still recommended for ideal loops.

Susume's picture
User offline. Last seen 2 days 19 hours ago. Offline
Joined: 10/02/2011
Better link for article

Here is a link for the full protein design article (author manuscript), which allows viewing the figures at a larger size:

Joined: 09/24/2012
Groups: Go Science
a question for fig 2

Fig 2 presents the following macro structure for Fig2a:
babbab = SHSSHS (S= sheet, H = helix)
or with the same number of 76 segments, the following Foldit SS:


If I understood well the paper, the length of the helixes and sheets and loops are "ideal" for this 76 segments case.

Do you think that any other (common by foldit players) macro structure like
SSSSHH or HSSSSH would have much lower chance to be stable? Or was the SHSSHS just given as example?

Joined: 09/24/2012
Groups: Go Science

Here are the SS for figure 2, starting from "N" to "C" in the figure (note that only few loop parts have >3 residues). I don't know if there is any needed direction from 1 to n or from n to 1.

n=76 Fold 1

n=99 Fold 2

n=70 Fold 3

n=100 Fold 4:

n=97 Fold 5:

Susume's picture
User offline. Last seen 2 days 19 hours ago. Offline
Joined: 10/02/2011
Other lengths can also be ideal

Other lengths for fig 2a can also be ideal - what needs to be controlled are 1) odd vs even numbers of sheet segs, and 2) length of helix is appropriate for length of sheets. The given example has sheets 7 segs long (odd). The same shape could be made with sheets 5, 9, 11 or 13 segs long (odd). The given helixes are 18 segs long - if you use 5 segs for sheets, the helices should be one turn shorter; for 9 segs they are probably long enough as is, and for 11 sheet segs the helices would need to be one turn longer. All the other examples in the figure can also be ideal with longer or shorter structures.

Personally I have not been able to make 18 segs of helix work for 7-seg sheets; I much prefer 17 segs. In general I use 17 or 21 where they use 18, and 13 where they use 14. What is important is that the end of the helix is pointing the right way to place the next sheet strand in the sheet plane.

For each of the given examples, there is a similar fold that runs the opposite direction (put seg 1 where last seg is and vice versa). In this case any sheets with a sheet-sheet turn between them need to be one seg shorter or longer (even number) because the ideal sheet-sheet loop is going the opposite direction. In some cases I find one of the helices also needs to be one seg shorter - this may depend which ideal loop shape you choose.

I have made several of fig 2a and fig 2c with different length structures, with different ideal loops, and in opposite direction, all getting full filter points. I have also omitted or added one sheet on each of these, in various places, to make novel folds in different lengths. There are other folds possible as well using the same rules, and when you find one you can change its structure lengths, use different ideal loops, reverse direction, or add or omit a sheet to make more novel folds. There is no example given using a helix-helix ideal loop - there are many shapes possible that follow the ideal design rules and have two helices together plus some sheets. Plenty of room for creativity!

Susume's picture
User offline. Last seen 2 days 19 hours ago. Offline
Joined: 10/02/2011
helix lengths

Looking at the helix lengths, I see they prefer a 2-seg loop at one end of the helix and 3-seg at the other, while I generally use 3-seg loops at both ends - I think this explains why my helices are one seg shorter than theirs.

Susume's picture
User offline. Last seen 2 days 19 hours ago. Offline
Joined: 10/02/2011
"ideal" helix lengths changed in game

As of puzzle 1294 (Oct 2016) the Ideal SS tool in the game curls the helixes at a slightly different angle than it used to, such that the helix lengths given in the ideal design papers are correct in foldit as well. The reason I kept needing different length helices than the scientists used was because the Ideal SS tool was curling them a little different than what the scientists started from.

Joined: 09/24/2012
Groups: Go Science
Correcting a bad loop (example for sheet-sheet connection)

Do you think it's possible to correct a bad loop?

Examples for a two loop connection (LL):

1) Rule 1: both sidechains of ends sheets should point to the same direction (bellow or up).
If wrong, impossible to correct ?

2) Rule 2: sheets must turn left (end sidechains down)
Correction? Replace each end sheet segment by L (gives LLLL). As the end sidechains turn from up to down, it turns from right to left. LLLL is neutral for direction.
Additionally, we have to add a sheet segment on the other end in order to keep sheets odd/even.
OR insert 2 L segments between in order to make LLLL
Then we have to retrieve these segments elsewhere in the protein.

Is it true that there is no mean to correct a bad "left" rule 1?

Joined: 09/24/2012
Groups: Go Science
Foldit image for Fig 1a

"A sheet-sheet connection with 2-3 loops must turn left"
Fig1a Left rule sheet-sheet23Fig1a Left rule sheet-sheet23
I obtained this by
1) cut the loop
2) align the sheets for bonds, wiggle, uncut
3) mutate loops to glycine
4) adjusting with Rama map in selection interface (both glycines must be in green or yellow)
5) shake, wiggle (if rama not ok, undo)
4) filter on for ideal loop (shows red bowls)
5) freeze unless ends of sheets
6) rebuilds, shake, wiggle until Filter AND Rama ok for the 4 residues

Joined: 09/24/2012
Groups: Go Science
oops ! correction ?

Correct me if I'm wrong. From Fig 1d, it seems that the direction of the last sidechain from the first sheet is not important (even if it's up, we are still turning left in the figure above). Then my "First rule" of "both to the same direction" would be erroneous. :P

Susume's picture
User offline. Last seen 2 days 19 hours ago. Offline
Joined: 10/02/2011
Possible to fix loop when sheets turn the wrong way

In this figure you have the sheet sidechains closest to the loop pointing away from you - and the loop turns left - this is correct. If the closest sheet sidechains were pointing toward us, the loop would have to turn right - it just means you are looking at the left-turning loop from the other side (I think of this as viewing from the "back").

If you placed your sheets so the closest sidechains faced away (so you are viewing from the "front") and the loop turned right, it would break the ideal design rule, and an ideal loop would be unlikely to form. The easiest way to fix this is to add one segment to each sheet next to the loop (between n and n+1, and between n+2 and n+3). Then the NEW sheet sidechains would face toward us, and the right-turning loop could be made ideal (really left turning, but you are now viewing from the back).

Another way to fix the above problem would be to steal one seg from each sheet and make it loop - then the new closest sheet sidechains would be facing toward us (the ones facing away have been changed to loop), and the 4-seg loop would turn right. If you turn the protein so the new closest sheet sidechains face away (to view from the front), you can see it is "really" a left-turning loop. There is one 4-seg ideal loop that passes the filter, and it turns left. It is red-red-red-green, so you would have to make the 4th loop seg glycine (so it can be green), set the first 3 loop segs to helix (remember red on the rama map is helix), and try rebuilding or remixing or rama map dragging until the loop turns to red-red-red-green - then it will be ideal.

According to the ideal design rules, a 4-seg loop can also turn right (when viewed from the front), but the devs have not yet given us a right-turning 4-seg ideal loop, so it would not pass the filter.

Joined: 09/26/2016
Groups: None
Woooow Awesome, thnks for the

Woooow Awesome, thnks for the video, I'm absolutely new to this (just finished the tutorials) and this series of Black Belt videos seem too cool, gonna watch them all. Thnks for the links

jeff101's picture
User offline. Last seen 7 hours 49 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
Black Belt Folding videos:

https://fold.it/portal/node/996236 has links to more Black Belt Folding videos.

Joined: 06/06/2013
Groups: Gargleblasters
hoping new blueprint tool will work

red red green doesn't mean anything to me. I don't know how to mutate to that. Hopefully the blueprint tool will solve this problem.


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