Hi Antony, (Sorry for replying with Phenix specific suggestions in non-Phenix forum. Since the whole conversation started here and seems to move on I thought it's best to keep it here rather than move off-list or to a more tailored list. Also several people expressed the interest asking directly.)
ok, here is the list. Some options are not exactly what you want but relevant and (very) close, some are supposed to do just what you want. Also, I assume you have the latest version of Phenix. *1) Use phenix.real_space_refine:* http://phenix-online.org/documentation/reference/real_space_refine.html For this you need a model (PDB file) and a map in CCP4 format or Fourier map coefficients. Since I'm not certain which one is going to work best in your specific case, I would run it in two ways (sorry, not GUI yet): phenix.real_space_refine model.pdb map_coefficients.mtz phenix.real_space_refine model.pdb map_coefficients.mtz run=minimization_global+local_grid_search Both should tidy up geometry while keeping best fit to the map, and also result in models with NO rotamer outliers or just a few (if you get outliers I would be interested to know, as this is not expected). The second option may run considerably longer. Resulting model should be a good starting point for usual reciprocal-space refinement in phenix.refine or other refinement program of your choice. map_coefficients.mtz can be MTZ file out of a phenix.refine (it will contain 2mFo-DFc map coefficients) or any other MTZ file containing an equivalent map coefficients. You may be prompted to specify them when running phenix.real_space_refine. Also, you can calculate the map like this phenix.maps model.pdb data.mtz or using the GUI. *2) Idealize model geometry minimally moving the model from starting (initial) position.* This is done using phenix.geometry_minimization (available in the GUI; some command line defaults may be different in the GUI): http://phenix-online.org/documentation/reference/geometry_minimization.html Using as simple as phenix.geometry_minimization model.pdb or drop in your file into the GUI, make sure you check "fix rotamers" box, and run. By default it will idealize all usual targets: bond, angle, planarity, dihedral, chirality, as well as it will use C-beta deviations restraints and for each residue side chain rotamer outlier it will switch it to the nearest valid rotamer. On top of that, optionally, you can add secondary-structure and Ramachandran plot restrains. Ramachandran plot restrains come in two different flavors, which is a topic of a separate discussion - let me know if you get to this! Important: since diffraction data is not used here, the resulting model can (potentially) deviate quite a lot from the starting one. Most likely you don't want this if you want to use this model to continue refinement. To make sure the model does not deviate much from initial state you can restrain it to starting point (I don't remember if these parameters are exposed in the GUI): phenix.geometry_minimization model.pdb reference_restraints.restrain_starting_coord_selection=all reference_restraints.coordinate_sigma=0.7 You can be more specific, and restrain only selected atoms to initial position: phenix.geometry_minimization model.pdb reference_restraints.restrain_starting_coord_selection="chain A and resseq 1:123" reference_restraints.coordinate_sigma=0.7 *3) Use phenix.refine.* In theory a default phenix.refine run includes step #1 above. In reality I'm yet to optimize how real- and reciprocal-space refinements play together in most optimal way. This means that while real-space refinement part of phenix.refine will fix all rotamer outliers, its reciprocal-space refinement part my not keep them all. Also, in case of large molecule real-space refinement in phenix.refine may take quite some time. *In summary*, my suggestion is to try all three options with all plausible to your case variants each one offers, and see which one works best. Good luck! Pavel On Wed, Jun 18, 2014 at 9:21 AM, Antony Oliver <antony.oli...@sussex.ac.uk> wrote: > Hi Pavel, > > Sorry… the current ‘triumvirate’ is, in no particular order: > > CCP4, Phenix and Buster (Global Phasing). > > Any suggestions would indeed be useful. > > Many thanks, > > Antony. > > - - - - - - - - - - - - - - - - - - > Dr Antony W Oliver > Senior Research Fellow > CR-UK DNA Repair Enzymes Group > Genome Damage and Stability Centre > Science Park Road > University of Sussex > Falmer, Brighton, BN1 9RQ > - - - - - - - - - - - - - - - - - - > email: antony.oli...@sussex.ac.uk > > tel (office): +44 (0)1273 678349 > tel (lab): +44 (0)1273 677512 > > http://www.sussex.ac.uk/lifesci/oliverlab > http://tinyurl.com/aw-oliver > - - - - - - - - - - - - - - - - - - > > On 18 Jun 2014, at 17:13, Pavel Afonine <pafon...@gmail.com> wrote: > > Hi Antony, > > Apologies for the cross-posting, but I *do* routinely use programs from >> all three software packages. >> >> I find myself refining a relatively low resolution structure (3.5 >> Angstrom) - with 8 molecules in the asymmetric unit. >> Is there a *simple* automated way to place “optimal-fit to electron >> density" side-chain rotamers into my model? >> Preferably in an NCS-independant manner? >> > > naively assuming that one of the "three software packages" that you did > not mention by name is Phenix: > yes, you can do it in a number of different ways. Let me know if > interested and I will list all options. > > Pavel > > >
