MTZMNF (CCP4: Supported Program)
- Identify missing data entries in an MTZ file and replace with a Missing Number Flag (MNF).
In a typical series of diffraction experiments, not all Bragg
reflexions for a given resolution range are in fact recorded. Hence,
after TRUNCATE some reflexion data records may be entirely
missing from the MTZ file, although the reflexion indices lie within
the measured resolution range. It is strongly recommended that index
sets are made complete within the desired resolution range - a
script to do this is provided in $CETC/uniqueify. The MTZ
file will then contain records where there are indices but no measured
data. This means that it is easy to estimate completeness and later programs
can `restore' values if possible. Furthermore, a particular
reflexion may be recorded for the native protein but not for a
derivative, and the corresponding combined reflexion data record
should indicate `missing data' for the derivative.
To-date, missing data has been indicated in a variety of ways.
For example, a zero standard deviation is taken to mean that the
corresponding datum (e.g. structure factor amplitude) is missing.
In all cases, however, the indicator is a number upon which
arithmetic operations can (erroneously) be performed. This convention
has now been discarded in favour of representing missing data by
Missing Number Flags (MNF), which by default take the value
of an IEEE NaN or VMS Rop. All relevant
programs check for the presence of MNFs in input MTZ
files, and take appropriate action. In particular, when
displaying MTZ files using the program MTZDUMP (or the
script $CETC/mtzdmp) missing data can be identified and
are subsequently represented in the output in an unambiguous
All programs will now output MNFs where appropriate.
Where such values occur in an input MTZ file, they will
be carried through to the output. Alternatively, MNFs
may be generated when for some reason no value can
be calculated for a particular reflection and column.
The program MTZMNF has been provided to convert
old-style MTZ files to the new convention. The program relies
on being able to identify `missing data', and to this end
a number of cases are checked. These cases are explained
in detail in the section PROGRAM FUNCTION below.
When a missing datum is identified, the corresponding
entry in the MTZ file is replaced by a MNF. The value of
the MNF is taken from the header of the MTZ file, and will
typically be a NaN or Rop. For old MTZ files, which have no MNF
specified in the header, the MNF is automatically set to
NaN or Rop.
As a safety feature, only columns which are explicitly
specified with the LABIN keyword are converted. Columns which
are not specified via the LABIN keyword are written unchanged
to HKLOUT. Old-style MTZ files may still be used with all CCP4
programs, and old-style checks on missing data remain
in place (occurring after the check for a MNF).
However, new data sets, completed with $CETC/uniqueify
and combined with CAD, should automatically
include the necessary MNF entries.
The various data control lines are identified by keywords,
those available being:
LABIN <program label>=<file label>
A line giving the labels of the input columns from HKLIN
to be converted. Only the columns specified will be
converted to the MNF format; the remainder are output
unchanged. The allowed program labels
are Fi SIGFi Di SIGDi (i=1,20), FCi PHICi (i=1,5),
Ii SIGIi (i=1,5), PHIBi FOMi HLAi HLBi HLCi HLDi (i=1,5).
The numbering used must be consistent. In particular,
if Di and SIGDi are specified then Fi and SIGFi must
also be specified, and must refer to the structure factor
amplitude associated with the anomalous difference data,
see example below. Futhermore Fi/SIGFi, Di/SIGDi,
FCi/PHICi, Ii/SIGIi and PHIBi/FOMi must be specified
in pairs, e.g. it is an error to specify F1 but not
SIGF1. Columns for which an appropriate program label is
not supplied by the program cannot be converted. This is usually
because an appropriate conversion protocol does not exist.
Note: The conversion of FCi/PHICi columns may in some
cases be dangerous - see the section PROGRAM FUNCTION below.
Title to be used in output log file and in output hkl file.
INPUT AND OUTPUT FILES
The input files are:
The control data file.
Reflection data file in MTZ format, assigned to HKLIN.
The output file is a reflection data file in MTZ format.
The printer output first gives details taken from the
input control data. Then header information from the input
MTZ file is echoed. Finally, a summary of the changes
made, i.e. the number of extra MNFs written to each column
specified in LABIN, is given.
The specified columns of HKLIN are assumed to fall into the
following groups: (1) Fi, SIGFi together with Di, SIGDi
if present; (2) FCi, PHICi; (3) Ii, SIGIi; (4) PHIBi, FOMi
together with HLAi, HLBi, HLCi, HLDi if present. For each
reflexion, each specified column is first checked to see if
a MNF is already present. If a MNF is found for one
member of a group, then all remaining members of that
group are assumed to be missing and are replaced by
MNF, with the exception that a missing Di/SIGDi does
not imply missing Fi/SIGFi. Next, an attempt is made
to identify `missing data' in the specified columns with the
If SIGF = 0.0 then SIGF and the corresponding F
(and D/SIGD if present) are replaced by MNFs.
If SIGD = 0.0 and the reflection is acentric
then SIGD and the corresponding D are replaced by MNFs.
However, if the reflection is centric
then SIGD and the corresponding D are not replaced.
If the calculated structure factor FC = 0.0
then FC and PHIC are replaced by MNFs.
WARNING: this situation is likely to be rare and
may even be dangerous! In exceptional cases, NCS
may allow a legitimate value of 0.0 to be calculated
for FC. On the other hand, FC = 0.0 may indicate use
of the wrong space group. Finally, low precision data
may cause a small but non-zero value of FC to be
confused with 0.0. The latter should not occur with
output from CCP4 programs, but may occur if FC data
If SIGI = 0.0 then SIGI and the corresponding I are replaced by MNFs.
If the weight FOM = 0.0 then PHIB, FOM and the
corresponding Hendrickson-Lattman coefficients HLA, HLB,
HLC, HLD (if present) are replaced by MNFs.
mtzmnf hklin $CEXAM/toxd/toxd_old.mtz hklout $CCP4_SCR/toxd_mnf.mtz
LABI F1=FTOXD3 SIGF1=SIGFTOXD3 -
D2=ANAU20 SIGD2=SIGANAU20 -
F2=FAU20 SIGF2=SIGFAU20 -
F3=FMM11 SIGF3=SIGFMM11 -
Martyn Winn, Daresbury
Eleanor Dodson, York University