Hydrogen Bond Network Filter

Hey everyone!

This post previews an upcoming Bonus Filter in Foldit design puzzles - the Hydrogen Bond Network Filter.

We are introducing this filter to overcome a major challenge in protein interface design, and something we've observed in many Foldit designs.

A lot of designs we've seen so far have used lots of hydrophobic residues in their interfaces. This works because these hydrophobic residues are buried in the core of the symmetric complex, sandwiched between two copies of the protein.

However, when you are designing proteins with interfaces, you have to consider that placing hydrophobics on the interface means that these residues will be on the surface of each of the individual pieces. This is a problem, because it means that these hydrophobics are exposed in each of the isolated pieces, making each piece unlikely to fold up correctly on its own. If the pieces don't fold up on their own, they wont be around to interface with one another!

To combat this problem, protein designers in the Baker Lab have been using Hydrogen Bond Networks.

A Hydrogen Bond Network is a 'web' of hydrogen bonds that connects the sidechains of multiple residues. When constructed across protein interfaces, these networks help to make the interface more stable. You can see an example of such a network below:

And unlike a hydrophobic core, these networks can be made out of polar hydrophilic residues. Because of this, the network will work well as a surface for each piece - and will still work well as an interface between the pieces!

The other major advantage of Hydrogen Bond Networks is that they increase the specificity of the interface. Since the networks are very carefully joined together like a jigsaw puzzle, each piece will be most happy when it's allowed to network with the other pieces. This helps to ensure that the pieces of your protein will interface with each other like you intended!

Hydrogen Bonds

Hydrogen Bonds are an interaction between a donor and an acceptor. As the name suggests, the donor donates a hydrogen atom, and the acceptor accepts it.

In Foldit, you can see these donors and acceptors using certain view options that are available when you've enabled 'Show Advanced GUI' in General Options.

Using the Score/Hydro+CPK Color scheme will color donors blue and acceptors red, like you see in the example above.

Donors have a partial positive charge, and acceptors have a partial negative charge. This causes the donor and acceptor to be attracted. But the opposite charges are not the full story; the charges are in specific locations and directions, and it's the geometry of the h-bond that really dictates its strength.

To get a better idea of how to improve your geometry, you can use the Stick+polarH option in 'View Protein'. This view option will let you see the hydrogen atoms themselves, shown in white.

Hydrogen Bond Networks

When multiple Hydrogen Bonds connect across multiple sidechains, they form a Hydrogen Bond Network. To get credit for a network in Foldit, you must have at least 3 hydrogen bonds, and at least 1 hydrogen bond that connects across an interface (these criteria may change from puzzle to puzzle).

In puzzles where the Hydrogen Bond Network Filter is enabled, you can visualize them using the filter. To do so, open the drop-down Filter Panel beneath the score panel, and click the 'show' checkbox for the HBond Network filter.

Valid networks will show up as a web of blue bonds. Each valid network that you form will get a score, which is then added to your total score as a bonus.

But what makes a good Hydrogen Bond Network?

Well, first off, your hydrogen bonds have to be good! A weak hydrogen bond wont work to extend a network. Remember that the strength of a Hydrogen Bond depends on its geometry. The donor and acceptor must be at the correct distance and angle to be strong enough for a network. Bonds that are too weak will show up in red on the visualization.

Once you have a network of good bonds, the score of a Hydrogen Bond Network is evaluated as follows:

Score = SCORE_SCALE * percent_polars_satisfied * num_intermolecule_bonds

So what do these mean?

SCORE_SCALE - This is just a constant that we set on a per-puzzle basis to decide how much Hbond Networks are worth.

percent_polars_satisfied - A good hydrogen bond network will minimize the number of unsatisfied polar atoms. If there's a blue or red atom in your network, it should be bonded to something - in some cases, multiple times. You want your network to satisfy most or all of its polar atoms.

num_intermolecule_bonds - The more bonds that your network forms across the symmetric interfaces of your protein, the higher it will score!

You can see the stats for each individual network by hovering over one of the bonds of that network. This will also highlight all bonds of that network in yellow:

It's important to note that while Hydrogen Bond Networks are awesome for stabilizing the interface between your symmetric proteins, you should have some hydrophobics on the interface as well. The best solutions will have a balance! Ideally, you want a very connected, perfectly satisfied network, with tight hydrophobic packing around it.

To go along with the release of this filter, we've posted a new puzzle http://fold.it/portal/node/2000714. In this puzzle, the backbone is fixed - your job is to mutate and move the sidechains to form networks of hydrogen bonds. Some players have already created some amazing networks in the devprev version of this puzzle! We're very excited to see what you can do, too!

( Posted by  jflat06 117 1504  |  Thu, 04/30/2015 - 23:54  |  16 comments )
jeff101's picture
User offline. Last seen 35 min 28 sec ago. Offline
Joined: 04/20/2012
Groups: Go Science
Some references:

In trying to understand the above discussion,
I did a google image search for "sidechain hydrogen bonds".
I found the following page with a section about hydrogen bonds:


This section discusses the hydrogen bonds between
H's on N-H groups and O's on C=O groups on the protein backbone.
It also has the following figure
about hydrogen bonds between protein sidechains:


Also, the figure below shows a
hydrogen bond between a glutamate sidechain and a serine sidechain
as well as a
salt bridge between a glutamate sidechain and a lysine sidechain:


It would be helpful to gather more figures like these showing
specific examples of hydrogen bonds between pairs of sidechains.
If you make these figures with Foldit,
please list the Color Schemes (CPK, Score/Hydro+CPK, etc.) you use.
I get confused because in plain CPK,
all hydrogens are white while nitrogens are blue and oxygens are red.
Meanwhile, above you say that in Score/Hydro+CPK,
donors are blue while acceptors are red.

bkoep's picture
User offline. Last seen 1 hour 14 min ago. Offline
Joined: 11/15/2012
Groups: Foldit Staff
Good links

The second link in particular should be helpful for determining how many hydrogen bonds are required to completely satisfy each residue type. For example, note that the N of tryptophan can only make one H-bond, whereas the N of lysine can make three!

Susume's picture
User offline. Last seen 1 day 16 hours ago. Offline
Joined: 10/02/2011
Would it be possible to get a

Would it be possible to get a list of how many donors and acceptors each aa has in foldit? The picture (jeff's 2nd link) shows 3 bonds possible for serine, but in foldit 2 bonds per serine in a triangle satisfies 100%, so foldit must count serine as having 2 possible. Similarly, the picture shows 4 bonds possible for asparagine, but 2 per asparagine in a triangle satisfies 66.67%, so foldit must count them as having 3 possible.

spvincent's picture
User offline. Last seen 4 hours 27 min ago. Offline
Joined: 12/07/2007
Groups: Contenders
I was going to ask the same

I was going to ask the same question but see Susume's already asked it.

So why don't serine triangles score at 66.67%?

(here's another source: http://www.imgt.org/IMGTeducation/Aide-memoire/_UK/aminoacids/charge/ )

bkoep's picture
User offline. Last seen 1 hour 14 min ago. Offline
Joined: 11/15/2012
Groups: Foldit Staff
To clarify about CPK coloring:

In Foldit, all CPK modes (including Score/Hydro+CPK) color nitrogen atoms blue and oxygen atoms red. Imagining blue donors and red acceptors is a helpful simplification, but is not strictly correct. While nitrogen is usually a donor and oxygen is usually an acceptor, there are exceptions to this rule. For example, the link above shows that one N of histidine can accept an H-bond, whereas the O of serine can also donate an H-bond.

Joined: 09/24/2012
Groups: Go Science

I've seen in the vancomycine puzzle, I think, also green spots in adition to blue and red.

What are these?

jeff101's picture
User offline. Last seen 35 min 28 sec ago. Offline
Joined: 04/20/2012
Groups: Go Science
The Foldit wiki has similar information:
jeff101's picture
User offline. Last seen 35 min 28 sec ago. Offline
Joined: 04/20/2012
Groups: Go Science

How does proline's backbone nitrogen behave in H-bonding?
I think the N loses its H to form a peptide bond with
its neighbor, as in the yellow part of the figure at

Does this mean the remaining N will be a hydrogen bond acceptor?

brow42's picture
User offline. Last seen 19 weeks 3 days ago. Offline
Joined: 09/19/2011
Groups: None
Proline N

The backbone N in a proline loses its ability to make an H bond.

Joined: 10/29/2008
Does resonance at the peptide

Does resonance at the peptide bond account for it being a weak acceptor, since it has some N+ character?

Joined: 10/29/2008

Does resonance at the peptide bond account for it being a weak acceptor, since it has some N+ character?

Joined: 10/29/2008

Those were in reply to sboyken below

sboyken's picture
User offline. Last seen 4 years 45 weeks ago. Offline
Joined: 09/01/2011
Groups: None
Yes, that is a good point and

Yes, that is a good point and is definitely part of it. Resonance is when covalent bonds, and the electrons that form them are in an equilibrium, alternating, between multiple states. Most bonds have some resonance, including peptide bonds (electrons are not static, far from it! we just draw them that way because it's easier):
Figure 2

As blivens correctly pointed out, the resonance of peptide bonds gives the backbone N a partial positive character. The affect of this resonance will be largely dictated by how planar the C-N peptide bond is (the phi/psi angles), and proline has the interesting property that it can form cis omega angles, which affects the C(i-1)-N(i) bond where i is proline. These are good questions!

@jeff101, one cool case in nature where proline does form h-bonds as an acceptor is the polyproline helix: https://en.wikipedia.org/wiki/Polyproline_helix
It's a helix of all prolines. Because there are no h-bond donors, the hydrogen bonds are with water molecules and other protein atoms (that are not part of the proline helix).

sboyken's picture
User offline. Last seen 4 years 45 weeks ago. Offline
Joined: 09/01/2011
Groups: None
@jeff101, you are correct

@jeff101, you are correct that the proline (Pro) nitrogen (N) loses it's H when it forms a peptide bond (it actually has 1 less H than other amino acids to begin with) and it will be an acceptor.

A better "short answer" than my original one is that yes it is an acceptor, but it is weaker, and is not a common occurrence contributing to backbone-backbone hydrogen bonds in protein structure, in the way that alpha-helices and beta sheets have h-bonds between N-H...O=C

A more detailed chemistry answer
Nitrogen has a valence of 5 -- this means it has 5 electrons in its outer shell. The rule of thumb is that atoms want 8 to be "happy" and they achieve this by forming covalent bonds with other atoms. So in the case of the proline nitrogen, the N makes 3 bonds -- 2 when it forms peptide bonds on each side (and 1 to itself) -- each of these 3 bonds shares 1 electron with the Nitrogen, so 3+5=8 and the N is "happy". But the N still has a "lone pair" of electrons, meaning that 2 of the 8 electrons are not participating in a covalent bond. The "lone pair" is what makes the N an acceptor. The partial negative charge of the lone pair electrons can attract the partial positive charge of a polar hydrogen

jeff101's picture
User offline. Last seen 35 min 28 sec ago. Offline
Joined: 04/20/2012
Groups: Go Science
Proline N

Are you sure the backbone N on proline does not become
like the lower N on histidine in the figure below,
with a red arrow pointing to it,
making it a hydrogen bond acceptor?


bkoep's picture
User offline. Last seen 1 hour 14 min ago. Offline
Joined: 11/15/2012
Groups: Foldit Staff
Satisfying polar atoms with H-bonds

In some more recent puzzles, we've emphasized that all buried polar atoms should be satisfied by forming hydrogen bonds. In reality, nearly all polar atoms in a protein will form at least one hydrogen bond. Polar atoms on the protein surface simply make hydrogen bonds with water; polar atoms in the protein core will need to make hydrogen bonds with other parts of the protein.

For those interested, this paper by Fleming and Rose discusses in more detail the importance of satisfying hydrogen bonds:

Download links:
  Windows    OSX    Linux  
(10.12 or later)

Are you new to Foldit? Click here.

Are you a student? Click here.

Are you an educator? Click here.
Social Media

Other Games: Mozak
Only search fold.it
Recommend Foldit
User login
Top New Users

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