MADMRG: Conversion of MAD data to pseudo-SIR with anomalous scattering
MADMRG is a routine for compressing a MAD dataset into a nearly-equivalent MIR + anomalous scattering pseudo-dataset. This is useful for refinement of parameters describing the anomalously scattering atoms for several reasons, not the least of which is the very large improvement in speed of refinement. This fast refinement is critical for the SOLVE automated structure determination routine. Ordinarily you will not run MADMRG by itself, but rather as part of ANALYZE_MAD or an automated structure determination. This means that you usually will not need to be familiar with all the keywords for MADMRG.
MADMRG reads in measurements of Fbar and the anomalous differences DelAno at several wavelengths for each reflection. From this data and the known values of f' and f" for the anomalously scattering atoms at these wavelengths, the program estimates (1) the magnitude of the structure factor corresponding to all atoms except the anomalous scatterer (Fo), (2) the "isomorphous" difference that would be measured +/- the anomalous scatterer at a standard wavelength, and (3) the anomalous difference that would be measured at this standard wavelength. In this way, the MAD data is converted to a form identical to that used in the analysis of SIR+anomalous differences data.
MADMRG assumes that structure factor due to anomalous scatterer is not large compared to that due to all other atoms. In this case iso differences among various wavelengths are proportional to differences in (f+f') for the anomalous scatterer, and ano diffs at each wavelength are proportional to f". MADMRG scales all the ano diffs to a common wavelength, then averages them. It takes all the iso diffs (e.g., L3-L1, L3-L2, L1-L2), and scales each iso diff by: (f+f' at std wavelength)/(difference in f+f' at the 2 wavelengths) to obtain estimate of what would be measured for the structure factor amplitude due to the entire structure at the standard wavelength minus the structure factor amplitude of the entire structure without the anomalously scattering atoms. That is, it estimates, delta Fiso (+/- ano scatterer at the standard wavelength). Finally, MADMRG obtains estimates of what delta Fiso would be at each wavelength by scaling std Fiso by f+f' at that lambda. This allows the program to obtain estimates of Fo, the structure factor amplitude due to all non-anomalously scattering atoms from each value of (Fbar - Fiso at that lambda). These estimates of Fo are averaged.
The result of these manipulations is a pseudo-SIR+anomalous differences dataset. The "native" structure factor amplitude is Fo, the estimate of the structure factor amplitude due to all non-anomalously scattering atoms at the standard wavelength. The "derivative" structure factor amplitude is Fo plus the isomorphous difference, delta Fiso, corresponding to the contribution of the anomalously scattering atoms at the standard wavelength. The anomalous difference is the averaged anomalous difference, scaled to the value at the standard wavelength. Generally, the standard wavelength is chosen to be one well away from the absorption edge of the anomalously scattering atoms, so that f' is small or negligible. This is not essential, however.
NSHELLS n number of shells of resolution used to group data is
n(default=10)
INFILE xxx input file is xxx
OUTFILE xxx output file is xxx
mad_atom xx anomalously scattering atom is "xx".
lambda xx wavelength number (1,2,3...) for values to be entered
next of f' and f" and for column numbers
wavelength xx wavelength value for lambda xx (e.g., 0.9798)
fprimv_mad xx 1 real number for f' value for anomalously scattering
atom at the current wavelength. Wavelength is defined
by the most recent value of the keyword "LAMBDA". You
need f' and f" regardless of whether you input
mad_atom or aval_mad...
fprprv_mad xx 1 real number for f" value for anomalously scattering
atom at the current wavelength. Wavelength is defined
by the most recent value of the keyword "LAMBDA"
LABEL label for this wavelength
NCOLFBAR n Fbar for this wavelength
NCOLSFBAR n sigma of Fbar
NCOLDELF n Del F ano (Fplus - Fminus)
NCOLSDELF n sigma of del F ano
JSTD n wavelength ID for wavelength to be considered the STANDARD
[default=1]
FIXSCATTFACTORS Fix scattering factors at their input values
REFSCATTFACTORS refine scattering factors f' and f"
[default=fix (i.e., do not refine them)]
OUTPUT FILE: This a DORGBN-style output file containing 8 columns of data. They are:
To use this data as MIR + anomalous differences, simply use columns 1 and 2 as Fp (native F) and sigma; columns 3 and 4 as Fder (derivative F) and sigma; and columns 5 and 6 as delano and sigma.
The last two output columns are for use in drawing Patterson maps only. Column 7 is equal to column 3 - column 1, multiplied by a weighting factor. Column 8 is equal to column 5, multiplied by a weighting factor. These weighting factors are based on a Bayesian weighting scheme. The mean square iso difference in a range of resolution, less the mean square sigma of this, is a good estimate of the mean square true iso difference, Del**2, in this shell. The observed iso differences are weighted down by a factor equal to Del**2/(Del**2+sigma**2). If sigma is small relative to Del, the weight will be about 1. If sigma is big relative to Del, the difference will be weighted down. A similar approach is taken for the ano differences. Look at the Patterson maps you get with columns 7 and 8 and compare them to the ones you get using column 3 - column 1 and column 5.
Please note: The output of MADMRG is set up to be used as "Mock" native and derivative data. When you refine heavy atom parameters using this mock dataset, you must define a heavy atom type that has scattering factors identical to those you use in MADMRG at the "standard" wavelength. That is, if you define lambda 3 as "standard" in madmrg and f" at lambda 3 is 8.9933, then when you get to heavy atom refinement with routine HEAVY you will need to define an atom type "L3" (or something) that has all the right scattering factor information including an f" of 8.9933. Use the keyword NEWATOMTYPE in HEAVY to do this easily.
Also note: when you use this data in your MIR program, DO NOT refine an overall scale factor and B for the "derivative." The overall scale factor and B of the derivative relative to the (pseudo) native are absolutely perfect to start with (because of the way the derivative has been set up). In this package, use the flag "NOREFINESCALE" for the derivative.
The reason to use column 4 as sigma of Fder is that heavy atom refinement programs such as HEAVY assume that errors in Fp and Fder are independent. In this case they are not. Suppose you estimated the error in Fder by combining errors in Fo and deliso. Then your heavy atom refinement program would estimate the error in Fder-Fnat by combining the errors you give it for Fder (based on errors in deliso + Fo) and the errors you give it for Fnat (the error in Fo). The estimates of errors in Fder-Fnat would therefore contain the errors in Fo twice. If you use column 4 as sigma of Fder, the errors in Fder-Fnat will be correctly calculated based on deliso and Fo.
Input data file
This input data must be scaled carefully. MADMRG does not scale your data for you.
Number of protein residues and anomalous scatterers
The program assumes that the B-factor for the anomalously scattering atoms is similar to that for all other atoms. Using the number of protein residues and the number of anomalously scattering atoms on the next line, the program estimates the rms value of structure factor amplitudes due to anomalously scattering atoms as a function of resolution. Each of these are for the asymmetric unit.