In trimeric symmetry, credit not given for bond between the two symmetric copies

Case number:671071-2005682
Topic:Game: Other
Opened by:NinjaGreg
Opened on:Wednesday, September 5, 2018 - 04:55
Last modified:Wednesday, April 15, 2020 - 23:25

I made a network with three segments by mutating one hydrophilic residue to glutamine, then wiggling. It bonded to the symmetric copies of itself, acceptor-to-donor, around the symmetry point. But the network filter says it only gets credit for 66% polars! From the picture of the network, you can see that two of the blue bonds have white cores, but the third (between the two symmetric copies) does not.

Is this the intended behavior? What's the point of the trimer design if the two copies don't behave the same as the original?

(Wed, 09/05/2018 - 04:55  |  6 comments)

LociOiling's picture
User offline. Last seen 10 hours 38 min ago. Offline
Joined: 12/27/2012

We discussed this quite a bit in chat, and the scoring still doesn't make sense to me.

I think that the blue-and-white barber poles don't have anything to do with scoring. Foldit draws them only on the primary chain, and for any bonds that cross between the primary chain and one of the symmetry chains. Bonds crossing between two symmetry chains aren't shown.

On the scoring, my understanding is that for glutamine, each nitrogen and each oxygen at the tips can have two bonds. So 4 bonds per glutamine, for a total of 12 possible bonds in the picture here.

I assumed that each blue spiral counts for two bonds, one for each segment involved. So 3 blue lines here, 6 bonds out of 12 possible, for a 50% score. Obviously, this doesn't agree with what NinjaGreg is seeing. So we are officially confused.

LociOiling's picture
User offline. Last seen 10 hours 38 min ago. Offline
Joined: 12/27/2012

A couple of references regarding hydrogen bonds:

Hydrogen Bond Network Filter blog post from 2015

Sidechain Bonding Gallery on the wiki

LociOiling's picture
User offline. Last seen 10 hours 38 min ago. Offline
Joined: 12/27/2012

Just re-read the Puzzle 1570 objectives, which states in part:

Networks must be at least 75% satisfied (i.e. 75% of all polar atoms in a network must make a hydrogen bond).

So that just talks about atoms, nothing about the possible bonds per atom. The network in NinjaGreg's post has six polar atoms, and all are bonded. So it should be at 100%, assuming the other criteria are met.

The hover help for the Hydrogen Bond Network objective is more detailed:

Networks must have at least 3 Hbonds, 1 Interface HBonds, and satisfy at least 75% polars. You can get credit for up to 8 Interface bonds. Bonds must be between two core residues.

Again, the network pictured seems to meet these requirements.

So it seems like NinjaGreg is correct, it seems that only bonds involving the primary chain are being counted, and bonds between the symmetry chains are not.

bkoep's picture
User offline. Last seen 1 day 4 hours ago. Offline
Joined: 11/15/2012
Groups: Foldit Staff

Thanks for bringing this up, NinjaGreg and LociOiling!

This is not a bug; the HBond Network objective is behaving as intended. I think the confusion here is a result of poor communication on our part, and also some shortcomings of the H-bond visualization in Foldit.

Appropriately enough, the element that is missing in this discussion about hydrogen bonds is: hydrogen. If you go to View Options and select "Stick+polarH" in the View Protein menu, you should see that there are two white hydrogens branching off of the blue nitrogen atom of glutamine.

Both of these hydrogens count as polar atoms that need to be satisfied (not the blue nitrogen atom itself). Each of these hydrogens is an H-bond donor, and the red oxygen of glutamine is an H-bond acceptor, for a total of three polar atoms on each glutamine; and a total of nine polar atoms in the entire network. However, the bonds in this network only satisfy six of those polar atoms, leaving three hydrogens high and dry—this is why the objective reports "66% satisfied".

We can definitely improve the HBond Network objective to make this more clear in the future. At the very least, I think clicking "Show" for this objective should automatically show polar hydrogens (like the Stick+polarH view option). Maybe hydrogens should also be shown with the "Show bondable atoms" View Option.

NinjaGreg's picture
User offline. Last seen 2 days 16 min ago. Offline
Joined: 05/21/2010
Groups: Go Science

Yes, the "Show bondable atoms" option on the view menu really should show the two atoms on the Nitrogen, and one on the oxygen. If possible, the bond display should connect to only one of those, and any second band should connect to the other. Up until now I have found the Cartoon Thin representation to be of great help in abstracting the plethora of atoms down to understandable levels of complexity, but now we need to see and understand the details of bonding, since it will be a major emphasis going forward.

Thanks for the quick feedback, I'll let my team know.

LociOiling's picture
User offline. Last seen 10 hours 38 min ago. Offline
Joined: 12/27/2012

The attached files were extracted from ir_solution files on 20200409-d067627d3b-win_x86-devprev.

Even though the two solutions have the same pose, the hydrogen bond scoring is different.

(Edit: the solutions were from puzzle 1812. Using this thread as a roundabout way of sharing with scientists.)


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