REFMAC (CCP4: Supported Program)

User's manual for the program refmac_5.*

Keyworded input - Other Xray keywords

Anything input on a line after "!" or "#" is ignored and lines can be continued by using a minus (-) sign. The program only checks the first 4 characters of each keyword. The order of the cards is not important except that an END card must be last. Some keywords have various subsidiary keywords. The available keywords in this section are:

Number of the resolution bins
Limits of allowed B value range
Cell parameters
Factors to scale down shifts at every cycle
End of keywords
Flag of the reflections excluded from the refinement
Output MTZ labels. Useful for map calculation
Refinement mode
Level of monitoring statistics during refinement
Parameters for the phased refinement
Parameters of the rigid body refinement
For scaling of the external partial structure factors
Shannon factor to control grid spacings
Number of TLS cycles

BINS | RANGE <nbins> | <range>

Number of resolution bins <nbins>. Default 20, maximum allowed 100. For least-square refinements it is useful only for monitoring statistics on resolution ranges. For maximum likelhood it is used for normalisation to convert Fs to Es (i.e. normalisation coefficients are calculated in these resolution bins).

Or <range>. If the value after BINS/RANGE is less than 1.0 this is assumed to define the bin width in units of 4*sin**2/Lambda**2.

BLIM <Bmin> <Bmax>

[Default 2.0 500.0]
Bmin and Bmax are minimum and maximum B-factors allowed. The lower limit is required by the program, but is set by default.

CELL <a> <b> <c> [ <alpha> <beta> <gamma> ]

Defines parameters of the cell. If this keyword is specified then cell parameters from MTZ and coordinate files will be overridden. This keyword could be important when cell dimensions are suspect and the user wants to change them. This keyword is generally not recommended.

DAMP <Pdamp> <Bdamp>

[Default 1.0 1.0 for resolution > 2.5Å, 0.5 0.5 for resolutions < 2.5Å]
<Pdamp> <Bdamp> scales shifts after each cycle of refinement. If there is limited data, or you are using unrestrained refinement, it is sensible to scale down the shifts.

FREE <nfree_exclud>

[Default 0]
The default value for exclusion for the FreeR is zero. However there is an opportunity to reset this exclusion flag here. See the description of the program FREERFLAG.

LABOUT <program label>=<file label> ...

This keyword tells the program that in output MTZ file calculated structure factors, their phases, coefficients for map calculations should have given labels.

For example:

      #  Output labels for calculated structure factors and
      #  coefficients for weighted "difference" and "2Fo-Fc"
      #  maps. Corresponding phases will have labels PHWT, DELFWT
      #  figure of merit of phases will be FOM

      #  Another example when all labels are given explicitly
             PHDELWT=PH_fofc FOM=FOM_refmac

FP SIGFP (and FREE if available) plus the following additional columns are written to the output MTZ file. Labels can be assigned for:

By default, FP and SIGFP are scaled to match FC, i.e. in output file FP and SIGFP meant to be in the "absolute" scale. The overall isotropic B correction is applied to the FC terms. At the same time individual atomic B values are incremented by overall B value. Overall anisotropic B values are estimated in the first cycle and fixed for further refinement cycles.
FREE (if available)
Copied from input.
FC PHIC are the structure factor contributions from the input atoms only, without any FPARTi terms added.
FWT and PHWT are amplitude and phase for weighted "2Fo-Fc" map (2mFo-DFcalc).
DELFWT and PHDELWT are amplitude and phase for weighted 'difference' map (mFo-DFCalc).
PHCOMB is the combined phase to use with FP. It is output only if "experimental " or prior phase information were input.
If prior phase probability distribution is available (either through Hendrickson and Lattman coefficients or through centroid phases and their figure of merit) then all output Fourier coefficients except PHIC correspond to the combined phases.
FOM = <m> - The "figure of merit" for this reflection.

For the FWT and DELFWT terms, FP is scaled to be at the same scale as Fcalc,
FCalc= vector sum of {Dc FC + D1 FPART1 + D2 FPART2 + .. },
Di = <cos (2 pi s Delta x)> are the resolution dependent coefficient which reflects errors in the atomic parameters, and m is the figure of merit for this reflection. (PHDELWT and PHWT will be "combined" phases if HLA etc. or PHIB FOM have been assigned.) Otherwise they are equal to either PHIC or PHIC+180.

Missing Data: For those reflections where the FP are missing, mFo is set equal to dFc. Hence the terms become FWT=dFC and DELFWT=0.0. The m and D are based on the program's estimate of SigmaA.

Rebuilding into these 2mFo-DFcalc and mFo-DFcalc maps seems to be easier than using classic nFO-(n-1)FC and difference maps, consistent with the established technique for SigmaA style maps. One advantage here is that since the m and D values are based on the Free set of reflections they are less biased than the values obtained by the CCP4 version of SIGMAA after refinement.


[Default HKRF]


This invokes standard refinement of individual atomic parameters.
This invokes RIGID body refinement. The description of domains is given under keyword RIGID.
This invokes TLS refinement. Definition of rigid bodies is taken from TLSIN input file (refer to input output files).

MONI <subkeyword 1> [<subkeyword i> <value>]

This keyword controls level of monitoring statistics during refinement.

Default is:

MONItor DISTances 10.0
MONItor TORSions  10.0
MONItor ANGLes    10.0
MONItor CHIRals   10.0
MONItor PLANes    10.0
MONItor BFACtors  10.0
MONItor BSHPere   10.0
MONItor VDWrest    3.0
MONItor RBONd     10.0
MONItor NCSR      10.0

<subkeyword 1> can be one of:

means no monitoring. Program will work with very little output info.
means only overall statistics will be monitored.
means MEDIum number of statistics will be monitored, i.e. at the first and last cycles overall statistics and outliers will be output. In all other cycles only minimum information (R value, free R value, figure of merit and first two geometric statistics). The user can choose the quantity of outliers by choosing a number.
means overall statistics and outliers will be monitored at every cycle.

The subsequent <subkeyword i> are:

TORSION <badtor>
Print restrained torsion angles differing by more than badtor * torsig from the ideal value. Default is 10.
DISTance <baddis>
Print restrained bond distances differing by more than baddis * dissig from the ideal value. Default is 10.
ANGLe <bad_angle>
Print restrained bond angles differing by more than bad_angle *angle_sigma. angle_sigma comes from dictionary file. Default is bad_angle = 10.0.
PLANE < badpln>
Print restrained planes differing by more than badpln * plnsig from the ideal value. Default is 10.
VANderwaals (or VDWRestraint) <badvdw>
Print vdw and ionic contact distances that are closer than badvdw*vdw_sigma. Default is 3.0.
CHIRAL <badchi>
Print restrained chiral volumes differing by more than BADCHI * chisig from the ideal value. Default is 10.0.
BFACtor <bad_b>
Print out bonded atoms if difference between B values is more than bad_b *B_sigma. Default is 10.0.
BSPHere <bad_sphere>
Print out atoms if their anisotropy deviate from a sphere by bad_sphere * sphere_sigma. Default is 10.0.
RBONd <bad_rbond>
Print out atoms if rigid body restraint deviates from ideal by bad_rbond*rbond_sigma.
NCSR <bad_ncsr>
Print out atoms if either poistional parameters or B values deviate from the average value over ncs related atoms is more than bad_ncs*sigma. Default 10.0.

NCYC <ncycref>

[Default 5]
This keyword defines number of cycles of refinement.

PHASe SCBL <scblur> | BBLUr <bblur>

This keyword tells the program that probability distribution of given phase information should be altered (this can also be done through REFI PHAS). For example:

PHASe SCBLur 0.7 BBLUr 20.0

Program will apply blurring as follows:

HLAnew = HLA*scblur*exp(-(sin(theta)/lambda)**2*bblur)
HLBnew = HLB*scblur*exp(-(sin(theta)/lambda)**2*bblur)
HLCnew = HLC*scblur*exp(-(sin(theta)/lambda)**2*bblur)
HLDnew = HLD*scblur*exp(-(sin(theta)/lambda)**2*bblur)

or if PHASE and FOM are given: the program first generates HLA and HLB using the formula:

HLC = HLD = 0.

i.e. the Phase probability distribution is unimodal.


This keyword controls parameters of the rigid body refinement.

For example:

RIGIDbody NCYCle 10
RIGIDbody GROUp 1 FROM 1   A TO 100 A
RIGIDbody GROUp 1 FROM 200 A TO 300 A
RIGIDbody GROUp 2 FROM 101 A TO 199 A
RIGIDbody GROUp 3 FROM 1   B TO 500 B


GROUp <ngroup> FROM <resnum> <chnid> TO <resnum> <chnid> | EXCLude | TRANS | EULErangles


Domain number. Maximum number of domains in the current version of program is 100. Each domain can consist of 100 different pieces. For example:

RIGIDbody GROUP 1 FROM 10  A TO 100 A  #  A domain containing most of the A chain, and 
RIGIDbody GROUP 1 FROM 108 A TO 176 A  #  an embedded bit of a B chain, consisting of 
RIGIDbody GROUP 1 FROM 200 B TO 220 B  #  residues 200-220. 
RIGIDbody GROUP 2 FROM 10  B TO 100 B  # for example the equivalent domain across a NCS axis
RIGIDbody GROUP 2 FROM 108 B TO 176 B  
RIGIDbody GROUP 2 FROM 200 A TO 220 A
<resnum> <chnid>
Spans are specified by the residue number, and the chain ID. See examples. Default is: all atoms belong to one domain.
Exclude main chain, or side chain atoms from refinement. Rotation and translation are determined for the defined subset. Rotation and translation are applied for all atoms also. Default is NO EXCLUSION.
TRANS <Tx> <Ty> <Tz>
To add translation for this domain before starting refinement. Default is 0.0 0.0 0.0.
EULErangles <alpha> <beta> <gamma>
Rotate domain by these angles around its center of the mass (in degrees) before starting refinement Default is 0.0 0.0 0.0.
Print rotation angles (EULErangles) and rotation MATRix. Default is EULErangles, Rotation matrix and polar angles.
NCYCle <ncycle>
Number of cycles for rigid body refinement. Default is 40 cycles.

SCPArt < nsc1> <nsc2> <nsc3> ...

If NSCi are set, the FPARTnsci is scaled relative of the FC FC_tot = FC_calc(PHIcalc) + sc1*FPART1(PHIP1) + sc2*FPART2(PHI2) + ....

If NSCi set - that partial structure factor will be scaled

For example:

#   Scale only partial 1st, 3rd and 4th structure factors.
#   All others should be added to Fcalc without scaling
SCPArt 1 3 4

SYMM <symmetry>

Defines space group symmetry name or number. If MTZ or coordinate files have symmetry then they will be used. It is good idea to have symmetry in the MTZ and coordinate files. This keyword is not recommended.

TLSC <ncycle>

It defines number of TLS refinement cycles. If this keyword is specified then program will do TLS refinement only. It is useful if one is interested in TLS parameters only. If TLS refinement is considered as precursor of individual atomic refinement then


should be used. In any case it seems to be better if B values of all atoms are set to some predefined value before tls refinement using

BFACtor SET <value>

NCYCle <ncycle>


NCYCle 5

Number of cycles for idealisation or restrained or unrestrained refinement.

SHANnon_factor <shannon_factor>

This keywords tells the program to change grid spacing by a given factor.


SHANnon_factor 1.5

If this keyword is given then grid spacing for structure factor, gradient and second derivative calculation will change accordingly. According to Niquist for a given resolution if grid spacing is equal to grid_spacing_min=0.5 d_max, where d_max is maximum resolution in angstroms, then a discrete Fourier transform will not lose any information. It is true when structure factors are calculated from the map and vice versa. When maps are calculated from atomic model (or gradients and second derivatives are calculated using convolution of derivatives of atom with difference and "Hessian" maps), then finer grid might be needed. The reason is for example if grid spacing is coarse and atoms have small B values, then values of electron density at the grid points may not approximate atoms correctly. Default (1.5) is good compromise. But if desired it could be changed. Shannon factor tells the program that actual grid spacing should be equal to grid_spacing_min/shannon_factor. If <shannon_factor> is increased then calculation will require larger memory and more time. If it is too small then approximation will not be correct and the program might become unstable.


End of keywords. Time to do work.