Hi I am trying to refine the model in low resolution and there are multimer in the asu. I did rigid body refinement after MR , molecule did not move much and R/Rfree stays 45 % and If I do the overlap between the monomers, rms is 0.06. If I do simulated annealing or any type of refinement (group_adp), Rfree is high and R goes down, but the if do the superposition of the monomer, rms is 1.5 to 1.8. So like to know what kind of refinement is can be done in very low resolution between 5 to 6 ang data? Is it any strict ncs could be possible in phenix that should not move the monomer too much? suggestions are appreciated rn
I am trying to refine the model in low resolution and there are multimer in the asu. I did rigid body refinement after MR , molecule did not move much and R/Rfree stays 45 %
Did you define multiple rigid groups for rigid body refinement? If not, then it's not too surprising that it didn't move much since the MR programs typically do a good job on rigid body refinement anyway. So, use "sites.rigid_body" keyword for this.
If I do simulated annealing or any type of refinement (group_adp), Rfree is high and R goes down, but the if do the superposition of the monomer, rms is 1.5 to 1.8.
Try SA in torsion angle space (main.simulated_annealing_torsion=true). At this resolution you may need to refine one ADP per group (say per residues or any bigger selected group of atoms). Also, use TLS. Use NCS if available, and you can defined it's strictness by adjusting appropriate weights. Pavel.
multimer in the asu.
I did rigid body refinement after MR , molecule did not move much and R/Rfree stays 45 % Did you define multiple rigid groups for rigid body refinement? If not,
thanks Pavel.
I did tight the restrain and R/Rfree drops to 37/41 % with rms for the monomer is around 0.12.
Also I did surprise to see the ranges of R factor while doing simulated annealing but final R-factor
goes high to 37/41 % ?
Is it anything wrong with convergence?
step= 50 temperature= 2284.4 rmsd=0.6256 r_work=0.3061 r_free=0.3564
step= 75 temperature= 2174.4 rmsd=0.7957 r_work=0.2978 r_free=0.3483
step= 100 temperature= 2064.4 rmsd=0.8910 r_work=0.2978 r_free=0.3681
step= 125 temperature= 1954.4 rmsd=0.9354 r_work=0.2977 r_free=0.3710
step= 150 temperature= 1844.4 rmsd=0.9684 r_work=0.2953 r_free=0.3793
step= 175 temperature= 1734.4 rmsd=0.9924 r_work=0.2928 r_free=0.3819
Is it possible to ncs averaging , especially multi-domain averaging in heterodimer molecule using phenix.
thanks
rn
________________________________
From: Pavel Afonine
(group_adp), Rfree is high and R goes down, but
the if do the superposition of the monomer, rms is 1.5 to 1.8.
Try SA in torsion angle space (main.simulated_annealing_torsion=true). At this resolution you may need to refine one ADP per group (say per residues or any bigger selected group of atoms). Also, use TLS. Use NCS if available, and you can defined it's strictness by adjusting appropriate weights. Pavel.
Hi,
I did tight the restrain and R/Rfree drops to 37/41 % with rms for the monomer is around 0.12.
this looks reasonable...
Also I did surprise to see the ranges of R factor while doing simulated annealing but final R-factor goes high to 37/41 % ?
Is it anything wrong with convergence?
step= 50 temperature= 2284.4 rmsd=0.6256 r_work=0.3061 r_free=0.3564 step= 75 temperature= 2174.4 rmsd=0.7957 r_work=0.2978 r_free=0.3483 step= 100 temperature= 2064.4 rmsd=0.8910 r_work=0.2978 r_free=0.3681 step= 125 temperature= 1954.4 rmsd=0.9354 r_work=0.2977 r_free=0.3710 step= 150 temperature= 1844.4 rmsd=0.9684 r_work=0.2953 r_free=0.3793 step= 175 temperature= 1734.4 rmsd=0.9924 r_work=0.2928 r_free=0.3819
Are you running torsion angle SA? I guess the numbers above indicate the overfitting. Can you try the same run with more tight geometry restraints (smaller value of wxc_scale parameter) - try several different values (smaller than the default one, 0.5 and choose the one that gives more reasonable Rfree-Rwork gap)? If you have NCS - use it. Pavel.
participants (2)
-
Pavel Afonine -
r n