phenix.real_space_refine: a tool for refinement a model against a map

Description

The program refines a model into a map. The map can be derived from X-ray or neutron crystallography, or Electron Microscopy, and its quality can vary from good to poor. The program aims at obtaining a model that fits the map as good as possible while possessing a meaningful geometry (no validation outliers, such as Ramachandran plot or rotamer outliers).

Contact author

For questions, bug reports, feature requests: Pavel Afonine (PAfonine@lbl.gov)

IMPORTANT: any bug or problem report should be sent after trying the latest nightly build of Phenix:

https://www.phenix-online.org/download/nightly_builds.cgi

Bug or problem reports should include at least 1) all input files used to run, 2) list of parameters that were customized, 3) error message. Please use file sharing tools such as Dropbox to send large files.

Features

GUI

Graphical User Interface is available.

Usage examples

  1. Running with default settings:

    phenix.real_space_refine model.pdb map.ccp4 resolution=4.2
    
    phenix.real_space_refine model.pdb map_coefficients.mtz
    

    This will do 5 macro-cycles of global real-space refinement with rotamer, Ramachandran plot and C-beta deviations restraints enabled. If NCS is present it will be used as constraints with NCS groups found by the program automatically. Note: if map is used then resolution needs to be provided.

  2. Rotamer, c-beta and Ramachandran plot restraints as well as using NCS constraints are enabled by default. To disable these restraints:

    phenix.real_space_refine model.pdb map.mtz ncs_constraints=False \
    rotamer_restraints=False ramachandran_restraints=False \
    c_beta_restraints=False
    
  3. Request output model to have bond and angle rmsd from ideal not greater than certain values:

    phenix.real_space_refine model.pdb map.mtz target_bonds_rmsd=0.01 \
    target_angles_rmsd=1.0
    
  4. Specify Fourier map coefficients to use:

    phenix.real_space_refine model.pdb map.mtz label='2FOFCWT,PH2FOFCWT'
    
  5. Run refinement using global minimization (default), local rotamer fitting, morphing and simulated annealing:

    phenix.real_space_refine model.pdb map.mtz \
    run=minimization_global+local_grid_search+morphing+simulated_annealing
    
  6. If PDB file contains unknown to Phenix ligand a ligand CIF file needs to be provided. Ligand CIF file can be obtained using one of available tools in Phenix (see documentation for more details). Once CIF file is available it can be used as following:

    phenix.real_space_refine model.pdb map.mtz ligands.cif
    
  7. Group ADP (B-factor) refinement. Currently only one option available: restrained group ADP refinement with two B-factors per residue (one for main and one for side chains). This is enabled by default. The command below will do B factor refinement only:

    phenix.real_space_refine model.pdb map.mtz run=adp
    

    To make use of multiple CPU for B factor refinement use nproc=:

    phenix.real_space_refine model.pdb map.ccp4 resolution=3.2 nproc=12
    

    Note: multiple CPUs will only be used for B-factor refinement (which is done by default).

  8. Running rigid-body refinement. Rigid body refinement can be run alone or in combination with other refinement strategies. This will run rigid-body refinement only:

    phenix.real_space_refine model.pdb map.mtz run=rigid_body rigid.eff
    

    where rigid.eff contains definitions of rigid body groups:

    refinement.rigid_body {
      group = chain A
      group = chain B or chain C
      group = chain Z
    }
    

    This will run rigid-body and individual coordinate refinement only:

    phenix.real_space_refine model.pdb map.mtz run=rigid_body+minimization_global
    
  9. In case NCS is present the program will attempt to determine NCS groups automatically and use to define NCS constraints. Automatically found definitions for NCS groups will be printed to the log. It is possible to provide NCS group definitions manually:

    phenix.real_space_refine model.pdb map.mtz ncs.eff
    

    where ncs.eff contains definitions of NCS groups:

    pdb_interpretation {
      ncs_group {
        reference = chain A
        selection = chain B
      }
      ncs_group {
        reference = chain C or chain D
        selection = chain X
        selection = chain Y
        selection = chain Z
      }
    }
    

    NCS group definitions can be obtained using GUI or command line tool:

    phenix.simple_ncs_from_pdb model.pdb
    

    NCS operators are refined if NCS constraints are used. To disable refinement of NCS operators use refine_ncs_operators=False.

  10. Use if secondary structure (SS) restraints is controlled by secondary_structure.enabled keyword:

    phenix.real_space_refine model.pdb map.mtz \
    secondary_structure.enabled=True
    

    If HELIX/SHEET records are present in PDB file they will be used to construct restraints. Otherwise one of Phenix tools will be run internally to derive secondary structure information from the input model. DNA/RNA specific restraints will be generated internally. Not all provided SS will be used by default: some H-bonds that exceed a certain threshold will be filtered out. To force using all SS annotations use remove_outliers=False keyword.

    phenix.secondary_structure_restraints can create SS annotations and restraints given a model file. This will output HELIX/SHEET records:

    phenix.secondary_structure_restraints model.pdb format=pdb
    

    and this will create a parameter file that defines SS restraints ready to use for refinement:

    phenix.secondary_structure_restraints model.pdb format=phenix
    

    None SS annotation programs can determine SS 100% reliably in all cases! SS annotation given to refinement will be enforced on the refined model. This means that any errors in HELIX/SHEET records or in SS restraints file will be propagated onto refined model. Therefore it is critical to make sure that provided SS annotations are as correct as possible. It is recommended to use one of available programs (such as phenix.secondary_structure_restraints) to obtain initial guess at SS annotation and then verify it manually to remove incorrect and add missing annotations.

  11. Weight between data and restraints is determined automatically to achieve specified rms deviations for covalent bonds and angles, which default to target_bonds_rmsd=0.01 and target_angles_rmsd=1.0. Alternatively, the value of the weight can be specified manually:

    phenix.real_space_refine model.pdb map.ccp4 resolution=4.2 weight=123.
    
  12. Number of refinement macro-cycles defaults to 5, which is sufficient in most cases. If model geometry or/and model-to-map fit is poor then using more macro-cycles may be helpful:

    phenix.real_space_refine model.pdb map.mtz macro_cycles=12
    
  13. It is possible to define a bond between any pair of atoms, any number of bonds. Likewise, an angle restraint can be defined between any triplet of atoms, any number of such restraints. Similarly, dihedral, planarity and parallelity restraints can be added. This example adds two custom bonds (one of which is between symmetry related atoms) and one angle restraints:

    geometry_restraints {
      edits {
        bond {
          atom_selection_1 = chain A and resseq 123 and name O
          atom_selection_2 = chain Z and resname LIG and name ND1
          symmetry_operation = -x,y,z
          distance_ideal = 1.4
          sigma = 0.025
        }
        bond {
          atom_selection_1 = chain B and resseq 321 and name O1
          atom_selection_2 = chain Z and resseq 234 and name O2
          distance_ideal = 1.3
          sigma = 0.03
        }
        angle {
          action = *add delete change
          atom_selection_1 = chain C and resseq 2 and name CA
          atom_selection_2 = chain X and resseq 4 and name HG
          atom_selection_3 = chain Z and resseq 8 and name O
          angle_ideal = 120.
          sigma = 5.
        }
      }
    }
    

The lines above need to be saved into a file, say edits.eff, and then given to refinement along with all other inputs:

phenix.real_space_refine model.pdb map.mtz edits.eff

To verify that these restraints (along with others) were actually used in refinement inspect .geo file that lists all restraints for all atoms.

The GUI provides a grafical way of defining custom restraints.

  1. Reference model restraints come in two flavors. One allows to restrain model to its initial coordinates:

    phenix.real_space_refine model.pdb map.mtz \
    reference_coordinate_restraints.enabled=true \
    reference_coordinate_restraints.selection="chain A and resseq 12:21" \
    reference_coordinate_restraints.sigma=0.05
    

    Here only residues from 12 to 21 in chain A will be restrained with the weight 1/0.05**2.

    Another allows to provide a reference model in one or several files. This may be a higher quility (resolution) model and it does not have to be idential to refining model (similarity 85% or better, which is a parameter). In this example two reference models are used: chain Z from reference_model_1.pdb file and chain C from reference_model_2.pdb file. Chain Z from the first file serves as a reference for chain A in refining model, and chain C from the second file serves as a reference for residued 123-321 in chain B of refining model:

    refinement {
      reference_model {
        enabled = True
        reference_group {
          reference = chain A
          selection = chain Z
          file_name = reference_model_1.pdb
        }
        reference_group {
          reference = chain B and resseq 123:321
          selection = chain C
          file_name = reference_model_2.pdb
        }
      }
    }
    

To use reference model restraints a parameter file (for example, called reference.eff) with lines like above needs to be created and provided as input:

phenix.real_space_refine model.pdb map.mtz reference.eff
  1. To see all parameters:

    phenix.real_space_refine -h
    

Notes

References