How SOLVE works
MAD structure determination and overall SOLVE operation
The real power of the SOLVE package is the automated solution of MAD and MIR datasets. For a MAD dataset, all you need to do is tell the program about your space group and cell dimensions, where the unscaled intensity data files are, and what the scattering factors are for the MAD dataset. The program scales the data using localscaling in SCALE_MAD , calculates difference Patterson maps, compresses the MAD data into SIRAS-like data using MADMRG, calculates an optimized Bayesian heavy-atom Patterson using MADBST, and iteratively builds up and scores potential heavy-atom solutions for the MAD dataset with the SOLVE routine. Potential solutions are scored on the basis of (1) agreement with the Patterson, (2) "free" difference Fouriers, (3) the presence of "solvent" and "protein" regions in a native electron density map, and (4) the figure of merit of phasing.
SOLVE can either look exhaustively for solutions (trying all possible additions/deletions/inversions etc) or else just follow the best solution and keep adding on to it. The default is to follow the best solution (keyword is "mediumquick"), and SOLVE keeps looking at seeds until this process leads to a result with a figure of merit > 0.5 and a Z-score over 10. If you set the "veryquick" keyword, SOLVE will just look at one seed this way.
Final phases for the top solutions are written out along with Hendrickson-Lattman coefficients for calculating maps and solvent flattening. SOLVE refines scattering factors for the MAD data after heavy-atom parameters are found by comparing the effective occupancies of sites when refined against each possible set of dispersive or anomalous differences. SOLVE calculates final phases using Bayesian correlated MAD phasing.
SAD structure determination
SAD data is single-wavelength data with anomalous differences. SOLVE treats SAD data as a native (the averaged F+ and F-) and anomalously-scattering derivative (F+, F-) where the f' value is zero and the f" value is non-zero. The heavy-atom sites are found using the anomalous differences and phasing uses just the anomalous differences. Scattering factors are not refined. SAD phases require solvent flattening or other density modification before they are useful. RESOLVE works very well for SAD data.
MIR structure determination
MIR structure determination is almost the same as for MAD structure determination. The program scales the data using localscaling in SCALE_NATIVE and SCALE_MIR , calculates difference Patterson maps, and iteratively builds up and scores potential heavy-atom solutions for the MAD dataset with the SOLVE routine.
Combinations of multiple MAD or MIR datasets
SOLVE can treat multiple MAD or MIR datasets by converting them into one super-dataset that uses each native dataset other than the first one as a pseudo- derivative with no heavy atom. Then it uses correlated bayesian phasing to take into account any non-isomorphisms among the different datasets. In this way you can combine several MAD datasets, several MIR datasets, or MAD and MIR datasets. See the command COMBINE.
Datasets with anomalously-scattering native
SOLVE treats native datasets with anomalously-scattering atoms as two datasets. You enter the same data twice, once as a "native", and again as a "derivative" for which you specify a new heavy atom type with a zero real part and a non-zero imaginary part.