Dear Yuri, Why do you think you need 36 reflections per atom when atoms with anisotropic B-factors only have 9 parameters? You can get away with much fewer in many cases especially if you have good restraints. As Ethan points out, a drop in R-free after adding many parameters may be misleading. Proper testing will give you a clearer example.
The Hamilton test in Ethan's paper is implemented in PDB_REDO (http://scripts.iucr.org/cgi-bin/paper?ba5174) and I had a quick look at some refinement statistics for structures with ~21 reflections/atom (like your case): according to PDB_REDO's strict criteria anisotropic B-factors are acceptable in two thirds of the cases. This was tested with Refmac on 285 PDB entries; ShelX's new restraints may well increase the success rate. HTH, Robbie Joosten Netherlands Cancer Institute www.cmbi.ru.nl/pdb_redo > -----Original Message----- > From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of > Yuri Pompeu > Sent: Monday, September 17, 2012 20:32 > To: CCP4BB@JISCMAIL.AC.UK > Subject: [ccp4bb] B-iso vs. B-aniso > > Dear community, > > The protein model I am refining has 400 amino acids (3320 atoms). > Some real quick calculations tell me that to properly refine it > anisotropically, I > would need 119,520 observations. Given my unit-cell dimension and space- > group it is equivalent to about a 1.24 A complete data set. > However, I have had a couple of cases where anisotropic B-factor refinement > significantly improved R-work and R-free, while maintaining a reasonable gap > for lower resolution models (1.4-1.5 A, around 70,000 reflections). What is > the proper way of modelling the B-factors? > Any thoughts and/or opinions from the community are welcome. > Cheers,