I have been working on refinements on a few enzyme-inhibitor structures
lately and noticed the discrapencies in REFMAC TLS refinements between
the new version 6.3.0 and older version 6.2.0. There are two molecules
par ASU. I simply define each molecule as a TLS group. The problem seems
not to be incidental. I have accumulated a few examples here, with each
pair of parallel runs under identical running conditions.
Structure A B C
Resolution 1.80 2.30 2.30
R/Rfree prior to TLS 0.209/0.256 0.216/0.287 0.217/0.279
R/Rfree w/ TLS(6.3.0) 0.529/0.541 0.247/0.297 0.245/0.296
R/Rfree w/ TLS(6.2.0) 0.188/0.225 0.206/0.262 0.206/0.255
RMSD Bond/Angle (6.3.0) 0.806/31.57 0.030/2.618 0.027/2.522
RMSD Bond/Angle (6.2.0) 0.020/1.979 0.013/2.122 0.013/2.089
The TLS refinements in version 6.3.0 result in higher R/Rfree and poor
geometries. In one case (A) it was failed entirely. The calculated TLS
tensors are quite different in two parallel refinements, with residual
atomic B values much lower in version 6.3.0 than those reported by
version 6.2.0.
I hope I can get help to resolve this problem. Currently I have to switch
back to the older version for the TLS refinement.
Regards,
Huiying
-----------
Huiying Li, Ph. D
Department of Molecular Biology and Biochemistry
Natural Sciences I, Rm 2443
University of California at Irvine
Irvine, CA 92697, USA
Tel: 949-824-4322(or -1953); Fax: 949-824-3280
email: h...@uci.edu