I wonder if naive protein structure can be predicted by using true distance and angle constraints.
If we can do that, how do we choose rosetta applications and parameters?
Hi, smlewis. Thanks for your reply.
I found that HybridizeMover requires a template. But I have no template for builting model.
I have tried to use abinitio-relax for structural prediction and added experimental constraint as follows:
AtomPair CB 63 CA 65 FLAT_HARMONIC 3.47 1.0 0.50
AtomPair CB 67 CB 69 FLAT_HARMONIC 6.47 1.0 0.50
Dihedral C 3 N 4 CA 4 C 4 CIRCULARHARMONIC 1.71 0.50
Dihedral N 4 CA 4 C 4 N 5 CIRCULARHARMONIC -2.33 0.50
Dihedral C 4 N 5 CA 5 C 5 CIRCULARHARMONIC -1.72 0.50
But the result is very different from the naive structure. I'm wondering what went wrong.
Any help would be appreciated.
There are a thousand places this could go wrong. I don't actually know what is wrong - particularly I cannot parse "the result is very different from the naive structure". If you mean naive as constraint-free: I would expect that the constrained result should be very different, that's the point of constraints. If you meant 'native', then I don't know why you don't have a template for hybridize.
If you are doing effectively ab initio with constraints, I would guess the constraints aren't strong enough during the early stages; perhaps they are not on at all if they are not active in the scorefunction. I can guess another dozen things that might be wrong but I don't have enough data to tell.
You should probably re-ask this in a new thread; it's easier to get attention to unanswered posts and I clearly don't know enough about what you are trying to do to help.
We have increased the weight of the constraint and this problem is solved.
Thank you very much!