[phenixbb] simulated annealing omit map

Dale Tronrud detBB at daletronrud.com
Fri Nov 14 14:06:30 PST 2014

Hash: SHA1

   Pavel has already responded to most of these points but I would
like to add a few notes.

   The fact that after you build a ligand you see that the density
greatly "improves" is quite scary to me.  This sounds more like the
map is echoing back your model and not confirming it.  I would delay
building the ligand until everything else in the model has been
corrected to get the most reliable and unbiased map possible and then
try to figure out what is there.  There is no reason to rush the
interpretation if there is more work to do on the back side of the

   The mistake you appear to be making is to ask the question "Does my
ligand fit the map?"  Even if something has bound, you don't know for
sure that it is your ligand.  The fact that you built your ligand into
the pocket and your stats improved does not mean that building
something else wouldn't have improved them more.  You can't simply see
density and assume it is density for your ligand.

   The fact that you have to contour at lower than usual levels
indicates that whatever is there is not at full occupancy.  This makes
identifying the true occupant of your site much more difficult.

   First you have to decide what possible things might be binding to
your protein and identify the most visible characteristics of each.
I'm used to working with chlorophyll so I'll use that as an example.
A chlorophyll is a big and flat dinner plate.  Unlike most other big
conjugated molecules, a porphyrin usually has a ligand that bind right
in the center.  While heme is also a common porphyrin the metal at its
center usually is much more electron rich, while the chlorophyll has a
"flag" stiking up from one of its ringlets. These are the sorts of
details you have to have in mind when looking at your density.

   Also remember that if whatever is in your binding pocket is there
only 60% of the time, something else is probably there 40% of the
time.  Your density will include that too.  It is very handy to have an
isomorphous apo crystal form so you can see what that 40% looks like.
 If you have that apo data you can get a clearer view of the
difference upon binding with an Fo-Fo map.

   Pavel has listed a number of types of maps to calculate that are
all good possibility but each has its drawbacks (which he listed).  I
am a big advocate of the "discovery map" because if it doesn't show
clearly that your ligand is the best interpretation than you shouldn't
be writing the paper to begin with.  Holding off on building the
ligand as long as possible will give the best map possible and you the
best shot at not making a mistake.

   On the other hand the discovery map is just a difference map and
suffers the same limitation Pavel mentioned about the limitations of
bulk solvent models.  One solution is an idea advocated by Andy
Karplus that he calls the "clean difference map".  This will sound odd
from coming from a person who has spent decades trying to convince
people to refine their models against all their data, but just because
you refine against all the data doesn't mean you have to include it
all in your map calculation.  A distinction between the bulk solvent
model which may be inappropriately flooding into your pocket and the
density of a true ligand is that the true ligand will have high
resolution features.  You can filter out the undesired bulk solvent
contribution by leaving out the low resolution reflections when
calculating your Fo-Fc map.  This is what Andy calls a "clean
difference map".  The clean difference map will have artifacts due to
the low resolution series termination ripples, but they will be
different than the artifacts due to the inappropriate bulk solvent
model in your pocket.  The features that are common to the two maps
should be unrelated to either set of artifacts and more likely due to
the true contents of the pocket.  Give it a shot, it can't hurt.  We
usually cut off at around 6A resolution because between 5 and 6 is
when the bulk solvent scattering really starts to kick in.

   You also asked if burning your crystals could improve the map.  If
you have more crystals I would recommend soaking them in a higher
concentration of ligand.  No amount of fancy data collection is going
to make the density of a low occupancy compound stronger.  Getting the
occupancy up will do you worlds of good.

Dale Tronrud

On 11/13/2014 10:40 PM, Ivan IVANOV wrote:
> Dear All,
> Would you mind if I disrupt your conversation about the omit map
> or composite omit map ?? I would like to share with you I'm in the
> same situation as Bishal. I want to see a ligand in the structure I
> have. I am using Phenix for refinement. When I do a molecular
> replacement with the model or part of the model, it looks like
> there is continues density of the ligand on 2Fo-Fc and Fo-Fc map at
> lower counter level.
> 1st QUESTION: Once we got the phases from the model which serves
> for MR, what is the counter level limit for a 2Fo-Fc and Fo-Fc maps
> we can look at ?? Long time ago I saw some papers in the literature
> telling about a sigma of 0.4-0.5 there is still useful information.
> (This first map after MR is not biased by ligand I want to see). 
> Further going on the structure refinement, at some point I add the 
> compound I poorly see and I do 3 different refinements !! One with 
> compound occupancy 0.8, another one with 0.7 and another one with
> 0.6. Then the density for the compound after refinement appears. At
> that point I am on the second option in Pavel's suggestions below -
> I get probably a biased positive density in the map. BUT, I would
> like to notice that my Fwork and Ffree go significantly down. 2nd
> QUESTION: Does it mean the compound is there if statistics do 
> down??
> About first Pavel's suggestion below when I am doing refinement in 
> absence of the ligand it is the same situation Fwork and Free go
> down and the ligand density it looks like to be still there and it
> is NOT completely masked by bulk solvent (but I have to go still to
> lower counter lever for the 2Fo-Fc map).
> Another point I would like to arise. I've done a composite omit
> map with Phenix (because as I told you I like Phenix). One thing it
> is worth noticing when you are in a hurry or you want to see quick
> result because you are not patient enough, do not use Phenix for
> omit maps (it is slow :-) :-) :-) ). Except that I like very much
> Phenix. In comparison I did also the same maps with CCP4 and at the
> end it does not seem to change that much the result when you
> compare both omit maps coming from Phenix or CCP4.
> About my case the omit map (no biased) gives the same quality of 
> density for the compound as 2Fo-Fc map after MR. I have to go in a
> lower counter lever to see something.
> 3rd QUESTION: What is the limit for a composite omit map to look at
> ?? 0.5-0.6 sigma ??
> 4th QUESTION: Do you think I can improve the quality of the
> density if I burn my crystals (collect several multiplicity) ?? or
> probably murge different crystals data??
> I like very much Dale's offer about the discovery map. I believe 
> there is something there.
> Sorry for the long mail,
> Kind regards, Ivan
> On Thu, 2014-11-13 at 08:38 -0800, Pavel Afonine wrote:
>> Hello Bishal,
>> as discussed before, computing ligand-omit map is tricky.
>> Quoting my previous email:
>> """ There are two commonly used options for omitting the ligand
>> in order to demonstrate its presence/absence in Fo-Fc OMIT map:
>> 1) Physically remove the ligand from PDB file. Then do some
>> refinement and calculate Fo-Fc map.
>> 2) Keep ligand in the file, set its occupancy to zero. Then,
>> again, do some refinement and calculate Fo-Fc map. In this case
>> you may want to ask refinement program to not move the ligand or
>> move it only a little.
>> Now, here is why these two options are poor and will not give you
>> what you want.
>> In the first case the bulk-solvent mask will be set in the
>> ligand region and therefore it will mask ligand density
>> (bulk-solvent will be filled into the ligand region). Depending
>> on the strength of ligand density it may be masked completely or
>> deteriorated.
>> If you follow the second option you will always get positive
>> density in ligand area. This density may correspond to
>> bulk-solvent, ligand or mixture of both. That is there will be no
>> simple way to differentiate whether this density arises from the
>> ligand or bulk-solvent. """
>> Dale Tronrud offered a great alternative option:
>> """ An alternative you might want to consider is what I call the
>>  "discovery map".  At some point in the refinement process there
>> was a map that convinced YOU that this ligand was present.  You
>> should be the hardest person to be convinced so that map will be
>> both an omit map (because the model had been refined without the
>> ligand prior to this) and clear enough to satisfy the reader. """
>> Now, recently I improved composite OMIT map calculation in
>> phenix.composite_omit_map
>> tool, see picture that illustrates how it works:
>> https://www.dropbox.com/s/dbp8e0348v5p78h/fig_6.png?dl=0
>> I think at present that's the best option to follow if for some
>> reason you cannot follow Dale's suggestion.
>> Pavel
>> On 11/13/14 6:58 AM, Singh, Bishal wrote:
>>> Hello everybody, I am generating figures showing the electron
>>> density map around the ligands. I deleted the ligands from
>>> final co-ordinate file and then performed refinement  with
>>> identical parameters as before while keeping simulated
>>> annealing=true. I shall be thankful if someone could suggest me
>>> whether I consider mFo-DFc map or 2mFo-DFc map or both. Kindly
>>> also tell me the minimum acceptable contour level required for
>>> showing difference map. All structures are at 1.4 - 2.0 
>>> angstrom resolutions.
>>> Regards, Bishal
>> _______________________________________________ phenixbb mailing
>> list phenixbb at phenix-online.org 
>> http://phenix-online.org/mailman/listinfo/phenixbb
> _______________________________________________ phenixbb mailing
> list phenixbb at phenix-online.org 
> http://phenix-online.org/mailman/listinfo/phenixbb
Version: GnuPG v2.0.22 (MingW32)


More information about the phenixbb mailing list