> One can only assume that there is 2 schools of thought -
> 1) water is part of the problem of model building
> 2) water is the solution to residual reduction.
I'm a bit of an atheist when it comes to all supernatural phenomenon,
whether deities, water molecules floating 4.5 Å in outer space,
too-perfect Rfree values, etc.
Lately I have been relying on coot and/or phenix to find waters
automatically, and then I subject them to various reality checks (coot is
very good for this) and try to hack out any deviants.
What I am beginning to realize is that there are often many solvent
features that cannot be explained away as noise (perhaps occurring at the
same place in two molecules in the asymmetric unit, large, well-defined
peaks, and so forth) and are more likely ions (Mg++, Na+, SO4--, etc).
Many of these I would have manually mis-assigned as waters.
Are there any methods that allow one to input such ions as possible
candidates? What would be ideal is to have a program that says something
like "The density peak at x,y,z, has a .0973 chance of being real. It is
only 2.2 Å from N7 of G12, precluding water. Based on the mother liquor
ingredient list you provided, there is a 0.75 chance that this is Mg++,
0.25 chance it is Na+ (or whatever), based on a
crystallographically-derived library of 40,000 structural variants of the
ribosome.
Bill
|