Hello:
some more on water picking...
After quite a bit of experimenting, I found this working the way I like
(implemented in phenix.refine):
1) peak at mFo-DFc map is higher than ~3sigma, and
2) peak center is within a hydrogen bond to another atom (water or
macromolecule), and
3) peak has approximately the same shape as a water molecule would have
at this resolution and local environment, and
4) peak at 2mFo-DFc is higher than ~1.5 sigma,
5) refine B-factor (and occupancy, resolution permitting) of water only,
6) apply criteria 3-4 to waters,
after a round of coordinate and B-factor refinement (whole structure),
check criteria (2-3-4) are still ok, and
7) apply B-factor filtering criteria to all waters,
8) keep cycling "1-8" until convergence (number of defined macro-cycles).
Of course I don't mention various technicalities related to partial
occupancies, alternative conformations, Hydrogen atoms on waters in case
of refinement against neutron data or ultra-high resolution X-ray data.
These are special cases that get special treatment.
There are a few other technical tricks to make this process robust and
efficient at high resolution, higher than ~1.2A or so.
All the parameters of the above protocol are available to adjust so one
can customize the procedure.
Also, the whole process of water picking can be relatively safely
enabled at earlier stages of refinement than it is typically done, IF
"fix_rotamers=true" option is used. In this case there is no danger that
a water will end up in an unoccupied yet side chain density, since when
the side chains move during refinement the water is treated as just a
density peak, making possible for a side chain run into water's atom.
I'm still playing with this and some summary is here (for those who is
really interested -:) ): http://cci.lbl.gov/~afonine/rsr.pdf
Regarding sigma cutoff: one of more ideas to try out is to lower sigma
cutoff criteria as refinement progresses and the structure improves.
Pavel.
On 4/17/10 5:47 AM, Sudhir Kumar wrote:
hi all
sorry for such a basic query, i'ld like to know what is the acceptable
sigma cut off for waters to be kept in a model if data is of about 1.6 A.
thanks in advance
Sudhir Kumar
Research Scholar
Structural Biology Laboratory
SLS, JNU,
New Delhi-110067
