[phenixbb] question about pick water in autobuild

Thomas C. Terwilliger terwilliger at lanl.gov
Tue Sep 8 08:55:33 PDT 2009

Hi Qun,

With your command you need to add one more keyword to get rid of all the


I'm sorry that doesn't happen automatically if you say place_waters=False
...I will make that happen in the future.

Note however that these waters should not interfere with building...as
waters are always deleted prior to building model.  I note also that your
command does not add new residues; it only adjusts the ones that are
there.  To build new model add:


All the best,
-Tom T

>> Hi,
>> from what you explain, the main issue I see here is that the model
building program does not treat water simply as density peaks, and so
removes them (waters) if necessary to build a protein chain instead.
>> Otherwise, it's probably a good idea in general to fill in all
>> places with some placeholders like water, or probably isolated N atoms
would work even better (if I recall correctly).
>> Pavel.
>> On 9/7/09 6:18 AM, crystallogrphy wrote:
>>> Hi, Pavel:
>>> The problem here is that autobuild placed waters into the densities
where they are the N-terminal residues. Without the waters here,
autobuild will build residues, not waters, to decrease R and Rfree with
less model bias. Now with water picking, autobuild does not build more
residues, just water.
>>> I still do not know why 'place_waters=False' cannot work. Tom has not
given me the answer yet.
>>> On Mon, Sep 7, 2009 at 1:45 AM, Pavel Afonine <PAfonine at lbl.gov
<mailto:PAfonine at lbl.gov>> wrote:
>>>     Hi,
>>>     I hope Tom gives you the answer about how to stop water picking in
>>>     From my side, I'm wondering why it bothers you (I'm not saying
that not working option is good)? Why don't you consider those
waters as a tool for improving your map? You are free to remove
any of them them later on. This can be viewed as a prototype of
what ARP/wARP does, or other "density growing" techniques... In
your case you just temporarily place waters where the model is
still absent, and when you build those missing places just
consider waters as density peaks (I hope model building programs
assume waters as just density peaks and remove waters if they are
on the macromolecule's way). On the other hand, model
>>>     incompleteness may make the map ugly even for relatively well
defined places.
>>>     Pavel.
>>>     On 9/6/09 7:38 AM, crystallogrphy wrote:
>>>>     Hi,
>>>>     I am building a model against a 2.1A data set using
>>>>     phenix.autobuild. Right now Rfree is about 35%. I do not want to
pick water in this stage because I see denisty at N-terminal of
my model which need to build more residues. However, the program
still picks water even after I say 'place_waters=False'. Does
anyone know why?
>>>>     phenix.autobuild data=exptl_fobs_phases_freeR_flags.mtz
>>>>     model=AutoBuild_run_4_/overall_best-090905a.pdb
>>>>     seq_file=../ztb-sd.seq refinement.place_waters=False
>>>>     refinement.refine_with_ncs=True nproc=2
>>>>     ------------------------------------------------------------------------
>>>>     phenixbb mailing list
>>>>     phenixbb at phenix-online.org <mailto:phenixbb at phenix-online.org>

More information about the phenixbb mailing list