[phenixbb] Confused with ANAISOU in PDB

[Pavel Afonine]

Hi Partha,

> At your leasure, if you could explain a bit about the water picking 
> parameters, that would be great. In my case, for 132 residues and 55% 
> solvent content, the default setting was picking 330 waters, which is 
> bit too much I felt. Did change the b_iso_min = 10 and b_iso_max = 40, 
> just to see what happens.. (because I thought for ordered water the 
> later should not be too high..) but I was not sure what the other 
> parameters mean. Like pick water above 3 sigma, remove below 0.9 sigma 
> as we are used to from Arp..

refinement {
 ordered_solvent {
   b_iso_min = 1
   b_iso_max = 50
   primary_map_cutoff = 3
   secondary_map_cutoff = 1
   min_solv_macromol_dist = 1.8
   max_solv_macromol_dist = 6
   min_solv_solv_dist = 1.8
 }
}

Here are the "most important":

1) Waters with refined B-factors beyond this interval are removed: 
(b_iso_min = 1,  b_iso_max = 50). I presume you do not refine the 
occupancies for water molecules; otherwise there are similar parameters 
for occupancies as well.

2) phenix.refine uses two maps to select prospective electron density 
peaks: mFo-DFc (primary map) and 2mFo-DFc (secondary map). 
primary_map_cutoff = 3 and secondary_map_cutoff = 1 define peak heights; 
default is --  a peak must be higher than 3 sigma for primary map and 
higher than 1 sigma for secondary map to be considered as a water 
pick-candidate. Raising these values will select stronger peaks and 
obviously will result in lesser number of found waters.

3) Only peak located within this distance interval from macromolecule 
are considered (min_solv_macromol_dist = 1.8, max_solv_macromol_dist = 
6). Some people believe that 6.0A is too generous. You may try to reduce 
it to lesser values like 3.5...4.5A. Again this will reduce the number 
of found waters.

4) Waters separated by less than min_solv_solv_dist = 1.8 are removed.
 
The options "2)" and "3)" can have the highest impact on the number of 
located water molecules. So, I would play with those first and monitor 
R/Rfree as well.

The other parameters have higher "expert level" and lesser impact on the 
outcome, so I don't explain them until requested.

Pavel.