Hi David,
I'm trying to use experimental phases derived from an cryoEM density map do carry out reciprocal space refinement in phenix.refine.
I used phenix.map_box to get the map coefficients, and then the CCP4 program chltofom (with the -colin-phifom flag) to create the HL coefficients from F/PHI computed from the electron density map.
1. If I ask for the mlhl target, I get quite odd target values, e.g. for a 2.9 Å data set:
| normalized target function (mlhl) (work): -5862.311069 | | target function (mlhl) not normalized (work): -6438670043.811632
and attempted refinements more or less destroy my protein and increases both the mlhl target value and the R-factors. Is there an example data set I could access to help me find out what I'm doing wrong? (Using the ml target, as suggested in DeMaio et al, Nature Methods 10:1102, 2013, seems to work well, and has a postive target function, but seems sub-optimal if phase information is available.)
this is odd.. Normally you should use (example) phenix.map_to_structure_factors map.mrc d_min=3.4 to convert map to structure factors. The MTZ file from the command above should be ready for phenix.refine. It contains "Fobs", free-R flags and HL coefficients that represent map phase information. MLHL target will be used in this case, it is implemented exactly as described in the original paper: Pannu, N. S., Murshudov, G. N., Dodson, E. J. & Read, R. J. (1998). Acta Cryst. D54, 1285–1294.
2. Is there a way to do a "simple" least-squares refinement, minimizing sum( w_i * |Fobs_i - Fcalc_i|**2 ) where both Fobs and Fcalc are treated as vectors? Does the ls target perhaps do this if use_experimental_phases=True?
No, this is not implemented. Pavel