Add Building Blocks to Blueprint Tool for loops turning 90 or 120 degrees

Case number:845813-2005628
Topic:Game: Tools
Opened by:jeff101
Opened on:Sunday, August 26, 2018 - 06:49
Last modified:Sunday, December 27, 2020 - 19:58

In recent Monomer Design Puzzles, I have been making
structures composed of loops and parallel B-strands
(see for details).
These structures are somewhat cylindrical, but they
have triangular or square cross-sections. I want the
loops between the B-strands to bend about 90 degrees
for square cross-sections and about 120 degrees for
triangular cross-sections. In some puzzles with the
Ideal Loops Filter, I have found loops with these odd
angles that also pass the Ideal Loops Filter. Does this
mean that some Ideal Loops can turn 90 or 120 degrees?
Are there certain sequences and sets of phi,psi angles
that occur often in nature and are particularly stable
yet form loops that turn 90 or 120 degrees?

I think most if not all of the Blueprint Tool's Building
Blocks for loops between sheets turn 180 degrees.
Are all of these Building Blocks Ideal Loops? Would it
be possible to add some Building Blocks for loops between
sheets that turn 90 or 120 degrees instead? I think having
such Building Blocks would facilitate designing proteins
like in that also pass
the Ideal Loops Filter and form reliably in the lab.


(Sun, 08/26/2018 - 06:49  |  9 comments)

jeff101's picture
User offline. Last seen 7 hours 29 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

How about providing Building Blocks for loops
between sheets that turn the chain by other
angles, like (360 degrees)/n, where n is an
integer? How about angles like 72, 60, 45,
40, 36, 30, 24, 20, 15, 10, 5, and even
0 degrees?

jeff101's picture
User offline. Last seen 7 hours 29 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
The paper LIN2015: "Control over overall 
shape and size in de novo designed proteins"
by Lin, Koga, Tatsumi-Koga, Liu, Clouser, 
Montelione, and Baker at
may be useful here. 

LIN2015's Fig.1 shows 4 "lego blocks" 
labeled A, B, E, or G taken from 
4 different parts of the Ramachandran Map 
with very specific phi,psi dihedral angles. 
These 4 "lego blocks" can pair 16 different 
ways, and each pair gives a net change theta 
in the chain direction and a net twist tau 
in the chain. Fig.1C lists the theta angles 
30, 50, 60, 70, 100, 110, 130, 140, 150, and 
170 degrees for the 16 pairs of the 4 "lego 
blocks". LIN2015 then goes on to discuss 
combinations with 3 or more of these "lego 
blocks", which seem to allow even greater 
variety in the net change in the chain 
direction and the net twist in the chain.
Are there analogs to theta and tau values
for groups of 3 or more "lego blocks"?
How about for single "lego blocks"?

I think it would be useful to have within
Foldit's Blueprint Tool not just the usual 
Ideal Loop Building Blocks but to also have 
the 4 single-residue "lego blocks" A, B, E, 
and G from LIN2015. Then we could make many 
different combinations of these 4 "lego 
blocks" and obtain reliably for each 
combination a different net change in the 
chain direction and a different net twist 
in the chain. It would also help if we could 
select groups of one or more of these "lego 
blocks" within the Blueprint Tool, and Foldit 
would list each group's overall theta and 
tau values.
jeff101's picture
User offline. Last seen 7 hours 29 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

Perhaps a single-residue "lego block" from the Rama Map
region for polyproline helices would help too.

jeff101's picture
User offline. Last seen 7 hours 29 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

Along these lines, having Building Blocks
for cis/trans peptide bonds as well as
various helix types (as in makehelix
would help.

jeff101's picture
User offline. Last seen 7 hours 29 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
It would also be good to have LUA commands
to detect or set cis/trans peptide bonds 
or other Building Blocks. One could imagine 
recipes giving messages like "3 Building 
Block constraints are still being enforced.
Please remove them to let your protein 
explore more different conformations."
jeff101's picture
User offline. Last seen 7 hours 29 min ago. Offline
Joined: 04/20/2012
Groups: Go Science

I run Rosetta@Home on some of my computers and have 
seen it running jobs lately with names starting like
"XW_JG_11222020_A_E5_BBEAA", so I searched for "XW_JG" at
and found many similar names like below:


In the above names, I think E5 stands for a 5-residue sheet,
H14 H15 & H16 are for 14- 15- & 16-residue helices, and the 
BBEAA BAAB GB BAB & other parts are for loops. When I've 
watched such runs on my computers, they had the structure
sheet-helix-sheet-sheet-helix-sheet, as the names above
would suggest. Are some of the above loops ideal? What
theta angles would they give for comparison with those
in Fig.1 of LIN2015? If any of the above loops are ideal,
would you please include them in the Building Blocks 
available for use with Foldit's Blueprint Tool?


bkoep's picture
User offline. Last seen 2 days 13 hours ago. Offline
Joined: 11/15/2012
Groups: Foldit Staff

I'm not sure who launched those BOINC jobs, but I can ask around! Protein designers in the Baker Lab will use all kinds of loops, depending on the project. If a design project can tolerate a lower success rate, then sometimes it might make sense to design less-than-ideal loop shapes to meet the needs of the project.

Note that, in reality, there's not a hard cutoff between "ideal" and "non-ideal" loops. The Lin et al. 2015 paper showed that loop shapes are not equally distributed in natural proteins; rather, some loops are more "popular" than others. The authors suppose that the more popular loop shapes might be inherently more stable (and evidence supports this).

We selected the top 20 or so most popular loop shapes for the Foldit Building Blocks. We thought these provided a reasonable diversity while staying safely among the most prolific loop shapes. We could dig deeper into less popular loop shapes and provide them as Building Blocks, but they would be less and less useful for designing super-stable idealized proteins.

jeff101's picture
User offline. Last seen 7 hours 29 min ago. Offline
Joined: 04/20/2012
Groups: Go Science
Below are some more Rosetta@Home jobs I've seen:

I think sometimes in Foldit I've seen odd loops (ones 
that aren't listed as Building Blocks) pass Foldit's 
Ideal Loops Filter. It would be useful to have a list 
of all these odd loops, even if they are less likely 
to occur than the ones we have Building Blocks for.
Susume's picture
User offline. Last seen 3 hours 40 min ago. Offline
Joined: 10/02/2011

Great sleuthing on the meaning of those strings, Jeff - I had not seen that succinct way of encoding designs before. The ones you listed look like they are keeping most of the loops constant while varying the one sheet-sheet loop in the middle (which by turning either left or right will tend to drive that particular fold to be either ferredoxin-like or reverse-ferredoxin-like). Several of those loops are listed in the 2nd ideal design paper (Lin et al), in figure S3 in the Supplementary info, so they have already been shown to be popular/ideal. A few of them are not listed there, particularly the sheet-sheet loops they seem to be purposely varying. 4-residue sheet-sheet loops were found to have about a 50% chance of turning left or turning right, so maybe someone is further digging in to which ones go left and which go right?


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