You are here

FastRelax results in bad structure

3 posts / 0 new
Last post
FastRelax results in bad structure
#1

I use FastRelax to do refinement for my pose. However, the structure becomes completely different and of bad quality after refinement. Is there something wrong with my relax file?  If not, why did this happen? I added the original  and refined pdb file.

The relax file is below:

switch:torsion
repeat 2
ramp_repack_min 0.02  0.01     1.0 100
ramp_repack_min 0.250 0.01     0.5 100
ramp_repack_min 0.550 0.01     0.1 100
ramp_repack_min 1     0.00001  0.1 100
accept_to_best
endrepeat

switch:cartesian
repeat 1
ramp_repack_min 0.02  0.01     1.0  50
ramp_repack_min 0.250 0.01     0.5  50
ramp_repack_min 0.550 0.01     0.1 100
ramp_repack_min 1     0.00001  0.1 200
accept_to_best
endrepeat
 

AttachmentSize
pose.pdb155.37 KB
pose_refined.pdb351.25 KB
Post Situation: 
Mon, 2021-04-26 08:39
Eden

Gromacs has a comedy page describing the term "blowing up". Your protein, to use this highly technical term, blew up.

In FastRelax, there are two ways to combat this if the starting pose is bad —such as one downloaded from Swiss-Model. This is because the starting sidechains are all unhappy.

scorefxn = pyrosetta.create_score_function('ref2015')
# fixed bb
movemap = pyrosetta.MoveMap()
movemap.set_bb(allow_bb=False)
movemap.set_chi(allow_chi=True)
relax = pyrosetta.rosetta.protocols.relax.FastRelax(scorefxn, 5)
relax.set_movemap(movemap)
relax.apply(pose)
# free BB
relax = pyrosetta.rosetta.protocols.relax.FastRelax(scorefxn, 15)
relax.apply(pose)

However, structures can blow up if a cartesian method is used with a non-cartesian one. For example:

scorefxn = pyrosetta.create_score_function('ref2015')
relax = pyrosetta.rosetta.protocols.relax.FastRelax(scorefxn, 5)
relax.cartesian(True)  # Boom
relax.minimize_bond_angles(True)   # Super boom
relax.minimize_bond_lengths(True)   # Hyper boom
relax.apply(pose)

scorefxn_cart = pyrosetta.create_score_function('ref2015_cart')
relax.set_scorefxn(scorefxn_cart)
relax.apply(pose) # no boom

 

Tue, 2021-04-27 07:53
matteoferla

Thanks for your detailed explaination.

Sun, 2021-06-13 00:14
Eden