On Wed, Jan 25, 2012 at 12:44 PM, Karolina Michalska wrote:

Information extracted from PDB file header:    program_name    : REFMAC    year            : None    r_work          : 0.15793    r_free          : 0.19845    high_resolution : 1.45    low_resolution  : 36.92    sigma_cutoff    : None    matthews_coeff  : None    solvent_cont    : None    TLS             : False (number of groups: 0)  After applying resolution and sigma cutoffs:    n_refl_cutoff : 26357    r_work_cutoff : 0.1450    r_free_cutoff : 0.1687

Can someone, please, explain why there is such a big difference in R values?  When I do refinement in phenix I get practically the same results that Refmac produces.

It could be twinning - the output should indicate this somewhere. What are the re-calculated r_work/r_free values without applying cutoffs? -Nat

Hi Karolina,

Information extracted from PDB file header: program_name : REFMAC year : None r_work : 0.15793 r_free : 0.19845 high_resolution : 1.45 low_resolution : 36.92 sigma_cutoff : None matthews_coeff : None solvent_cont : None TLS : False (number of groups: 0)

this is exactly what phenix.model_vs_data extracted from the header of input PDB file.

After applying resolution and sigma cutoffs: n_refl_cutoff : 26357 r_work_cutoff : 0.1450 r_free_cutoff : 0.1687

These numbers are actually calculated by phenix.model_vs_data (after applying all resolution and sigma cutoffs, if any were reported in input PDB file header). These numbers may be lower than resorted ones for a number of reasons, among which these are the most likely ones: 1) account for twinning (if any present); 2) better bulk-solvent and overall (anisotropic) scaling used in Phenix; 3) rejection of Fobs outliers.

When I do refinement in phenix I get practically the same results that Refmac produces.

What I wrote above is based on the amount of information you sent us. If you send me the input data and model files then I will be able to look into the details and explain what's causing the difference. Pavel