Suggestion for HBond network bonus

Case number:845813-2010483
Topic:Game: Tools
Opened by:Bruno Kestemont
Opened on:Friday, October 2, 2020 - 07:59
Last modified:Thursday, October 8, 2020 - 14:11

HBond network bonus is extremely difficult to obtain. I think it's partly due to the "all or nothing" behaviour of the bonus.

Either you are close to a bonus (e.g. with red bonds) and you (and your recipes) get nothing, either you gain a lot of points.

Once you have the network, it's relatively stable, because recipes won't broke it.

But the score doesn't make a difference between "close to network" and far away monomers. It's very likely that "almost network" are broken by recipes.

Would there be any mean to give some partial score if we are near to a HBond network ? This would help to stabilize "near" to HBond network solutions (and help refining up to the network)

(Fri, 10/02/2020 - 07:59  |  6 comments)

Joined: 12/06/2008
Groups: Contenders

Bruno, I would like to know what the difference is between the red, "almost" H-bond, and the blue "real" H-bond. I realize the description says a red bond is too weak to form a network, but I have had the same two or three residues in a protein change color from red to blue or vice-versa, when I did nothing directly to them.

bkoep's picture
User offline. Last seen 1 day 16 hours ago. Offline
Joined: 11/15/2012
Groups: Foldit Staff

That's correct: a red H-bond is simply less stable and is worth fewer points. We want avoid these "marginal" H-bonds in networks, so we exclude them from counting in the network.

H-bond scoring in Foldit is actually dependent on the environment of the H-bond, which could explain the strange switching between red and blue. An H-bond that is more buried is more stable, and will have a higher score.

If you change the environment around a H-bond (e.g. make a mutation that exposes your H-bond to solvent), then that can change the score of the H-bond and cause it to switch from blue to red.

Joined: 04/20/2012
Groups: Go Science
"An H-bond that is more buried is more stable, and will have a higher score."

Does this mean that, other things being equal, an H-bond between segments
with larger Packing or Hiding subscores will be stronger and thus also have
a larger Bonding subscore? What subscores best reflect how buried a segment
is? What subscores best reflect how exposed a segment is? Is Bonding the 
subscore that best reflects how strong an H-bond is?

If you could cite papers or web pages (perhaps in the online Rosetta 
documentation) about how hydrogen bond strengths are determined, I would
appreciate it.

Joined: 04/20/2012
Groups: Go Science

I found an interesting chart about H-bond donors and acceptors.
It includes the sulfurs on methionines and cysteines as H-bond
acceptors. Does Foldit let sulfur atoms participate in H-bonds?
Will it show such H-bonds? Will it give them nonzero Bonding

Joined: 04/20/2012
Groups: Go Science

I wonder if Boots' problem was from making networks that
included H-bonds between glutamine or asparagine sidechains.
shows that for these sidechains, the nitrogen can donate 
2 hydrogens for H-bonding while the oxygen can accept 2.
To me, this means it may be possible to form a double H-bond 
between a sidechain N and a sidechain O on glutamine or
asparagine. If this happens, the way Foldit displays H-bonds
will cause confusion. It seems to always show H-bonds as if
they directly connect the N and O atoms, even if the hydrogen
atoms are shown. This means that visually, a single H-bond 
and a double H-bond would look the same. Perhaps Boots' 
network kept changing color as the glutamine/asparagine 
sidechains in it changed from single H-bonded to double 
H-bonded. One fix for this would be to have Foldit show 
H-bonds as connecting hydrogen atoms to the acceptor atoms 
when the hydrogens are shown.

bkoep's picture
User offline. Last seen 1 day 16 hours ago. Offline
Joined: 11/15/2012
Groups: Foldit Staff

Hmm, I see what you mean, Bruno.

But I don't think we want to award a partial score for a partial network. A partial network confers zero stability to a protein design (in fact, if there are BUNS in the network, a partial network will hurt more than it helps).

It seems like maybe the Objectives should have two separate functions for recipes: GetBonus and GetProgress (or something similar). This has come up before, but it might be time to revisit the idea.


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