Re: [ccp4bb]: Dm/dmmulti and solvent flipping

[Kevin Cowtan <cowtan@ysbl.york.ac.uk>]

***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***

(Should probably send a copy of this to the bb, since it is a FAQ)

The solvent flipping question involves several issues:

1. Why doesn't 'dm' do solvent flipping any more?

2. Why doesn't 'dmmulti' do it?

Let me deal with these in turn:

1. 'dm' does not do solvent flipping any more because solvent flipping was 
superceded by the gamma correction. In Jan Pieter's papers he shows the 
absolute equivalence between the two techniques. My own work went on to 
extend this the gamma correction to histogram matching and averaging. So you 
can look at this in one of two  ways: 'dm' doea not do solvent flipping 
because it does something more general. Or 'dm' still does solvent flipping, 
but it does it under another name.

2. 'dmmulti' does not do either solvent flipping or the gamma correction 
because extending dm to multi-crystal problems was one level of complexity 
too many for a Fortran program. As a result, maintaining or updating 
'dmmulti' has proven not to be worthwhile. In addition, bias correction 
methods like flipping/gamma are less important when averaging is available, 
so the need is less.

I am tackling the complexity problem by the 'Clipper' C++ libraries. These 
were originally intended specifically to allow a replacement for dmmulti to 
be written, although are now being applied to automated map interpretation 
and refinement. The complexity reduction is about an order of magnitude - 
Clipper programs seem to take about 10% of the lines of code of their Fortran 
equivalents.

However 'pirate' -  the C++ dm/dmmulti replacement - will not do flipping 
either. In fact it doesn't identify a solvent region at all, but instead uses 
continuous empirical density statistics and a bias-free method for applying 
them.

Kevin

-- 
Protein Structure Group, University of York, York, YO10 5YW, UK