You may need to create a bonus of, say 250 points for each hbond to an oxygen on that backbone to get people to focus on that. Maybe more, you'd have to play with it. Then it wouldn't be the best protein around the poly-proline helix, that sort of encloses or sits next to it.
That's a good idea! I'll see if it can be done.
I don't know about others, but I've found it near impossible to first, make an hbond to the polypropylene, then second to keep it bonded once the bands are removed. It seems like the repulsion of the backbones is stronger than the bonding.
That's rather interesting. Could you share some solutions that had bonds before bands were removed? We'll look into what is causing repulsion.
On 1778, I've been relying on the prolines to spontaneously hydrogen-bond to nearby residues. Very few prolines have done so, and when they have, their Bonding subscores have been low, like 3-5.
Can other proline subscores besides Bonding contribute to the stability of our poly-proline strands? Clashing? Pairwise? Packing? Hiding? Ideality? Backbone? Sidechain? Reference? Don't we really just want the best total scores we can get for the prolines in our poly-proline strands? Also, which subscores best reflect hydrophobic interactions? How about pi-stacking between a proline and nearby aromatic sidechains like his, phe, pro, trp, or tyr? Which subscores best reflect that?
One bonus/filter/objective I think you need for these puzzles should count how many prolines our structures have. This bonus might help prolines not get mutated away by recipes.
Also, for general hydrogen-bonds, do some directions for the hydrogen-bonds work better than others? Like if you have an N-H hydrogen-bonded strongly to an O=C, do all 4 atoms roughly form a line? What about for an O-H hydrogen-bonded strongly to an O=C, a bit like in water? Are all 4 atoms roughly in a line?
Thanks!
Thanks for the feedback!
We are interested in solutions that have Hbonds but score poorly. Sharing these solutions would help us understand why satisfying these polar atoms causes a drop in score.
Foy hydrogen bonds, it is true that if the atoms are in line, the interaction is strongest for a given distance.
To elaborate a bit more:
It's true that all of the sub-scores contribute to the stability of the poly-proline helix (hydrophobic interactions are captured by Hiding and Packing subscores).
However, we are especially worried about unsatisfied polar atoms here, because they are poorly modeled in Foldit. An unsatisfied polar atom is essentially a deal-breaker for protein folding, but this is difficult to model with our score function. Pi-stacking is also poorly captured in Foldit—but we should note that proline won't really participate in pi-stacking anyway, since it is not aromatic.
Hydrogen bonds are certainly directional! It's more important that the the donor O-H or N-H points directly at the acceptor (i.e. 3 atoms in line, with H in the middle); the alignment of the 4th atom is less strict. For details about H-bond geometry in Foldit, see Figure 3 in this reference.
Residue Count (max +100)
Penalizes a residue count above 80 by 20 Foldit points per additional residue. The count is capped at 85.