Hello, I have buid a homology model and I would like to refine it based on relax protocol. When I don't add any restrain to the model, the side chain changed a lot including these around the active pocket. I found two "fix backbone option" , one is "constrain_relax_to_native_coords", the other is "constrain_relax_to_start_coords", is there any difference between these two? I found the same commend in the help files.
However, I want to just add some constrain to the backbone but not completely fix it, for example: 20 Kcal/mol/A^2. Or is it possible just fix the alpha carbone of the protein?
Thank you very much.
The difference between the native and start options is whether the constraints are based on the starting pose for the trajectory, or a user-supplied native pose.
It is possible to set up your own constraints manually; I believe the abrelax code is compatible with constraints. There is a file in the manual describing how to write your own constraint file to constrain the backbone atoms as you want. http://www.rosettacommons.org/manuals/archive/rosetta3.1_user_guide/file...
There's no clean way to translate kcal/mol/area into rosetta energy units - you'll just have to try different coefficients in the functional form for your constraint until you get the strength you want. I suggest making a graph of the function you choose and setting it so that it adds 5 energy units or more in the zone you want to be forbidden. In fullatom mode, a few REU/residue is expected for a well folded protein, so 5 should be plenty. I don't know what you'd need in centroid mode.
Thanks a lot for your kind reply. It would be a challenge for me to write the constrain by myself.
I would like to add some constrain to the alpha carbon(the position won't be changed a lot and on the other hand it could still let rosetta to refine other part of the protein) so that it could keep the structure much more like my initial homology model. Otherwise, the refined structure would change a lot after refinement which is not reliabe to some extent, especially the sidechain conformations.
If possilbe, would you please provide me some ready to used scrips? I would be very appreciated for this.
BTW, does the relax protocol support muti-CPU?
if I use the following command, does it really save the job time?
mpirun -np relax.release @flags
I don't have any scripts for the purpose, sorry! Maybe someone else will.
The relax executeable in 3.1 runs JD2 internally, so it will readily accept MPI. Remember that you'll have to recompile in MPI mode for this (scons bin mode=release extras=mpi).
MPI will not increase the speed of any given trajectory - it just organizes them across processors so you can run one MPI job on N processors instead of N processes in N separate directories. So, yes, it saves users' time, but it won't make an individual trajectory finish any faster.