[phenixbb] anomalous difference map

Randy J. Read rjr27 at cam.ac.uk
Sun Mar 27 02:40:35 PDT 2011

Since you have a molecular replacement model, the other option you should 
try is to ask Phaser to look for the anomalous scatterers with 
log-likelihood-gradient maps in the SAD likelihood target. In our 
experience, this gives significantly better signal-to-noise on average than 
simple model-phased anomalous difference Fouriers.

By default, Phaser will put in anomalous scatterers where there are peaks 
above 6 times the rms of the log-likelihood-gradient map, and it does this 
iteratively, i.e. putting the anomalous scatterers into the model makes the 
model and makes the log-likelihood-gradient map more sensitive, so it 
carries on doing rounds of this until there are no more peaks to interpret 
as anomalous scatterers. At the end, you'll have a flat 
log-likelihood-gradient map but you'll also have a list of sites that were 
all 6sigma in at least one of the maps. As a bonus, you'll also have phases 
and a map that combine the information from your molecular replacement 
model and the anomalous scattering.

You can do this either through AutoSol (if you did the molecular 
replacement with AutoMR you will have been given the option to press a 
button to run AutoSol, which is convenient) or directly through Phaser-EP. 
In each case you have to give a PDB file identified as a partial molecular 
replacement solution.

Let me know if you have any difficulty finding the right options.

Best wishes,

Randy Read

On Mar 26 2011, Jason wrote:

>On Sat, Mar 26, 2011 at 11:29 AM, Nathaniel Echols <nechols at lbl.gov> wrote:
>> On Fri, Mar 25, 2011 at 9:20 PM, Jason <phenix.upitt at gmail.com> wrote:
>> > 1) Synchrotron Fluorescence scan indicates the heavy atom definitely
>> > presents in my protein (my control is a second protein with the same
>> ligand
>> > showed no absorbance spectrum)
>> Keep in mind that a positive fluorescence scan does not necessarily
>> mean that the element of interest is bound and well-ordered in the
>> crystal - it is easy to get an excellent scan without seeing anything
>> in the maps later.
>Indeed, but hopefully it's not the case.
>> > When I load the mtz file to phenix.maps GUI, the mtz label pulldown 
>> > menu indicates 4 possible choices for the column to use:(1) IMEAN, 
>> > SIGIMEAN (2) I(+), sigI(+), I(-), sigI(-) merged (3) F(+), sigF(+), 
>> > F(-),
>> sigF(-)
>> > merged (4) F, sigF, Dano SigDano. I have tried the choices of (2) (3) 
>> > and (4). However, there is barely any anomalous signal at 4sigma, 
>> > which makes
>> me
>> > wondering if something is not right. The first thing coming to my 
>> > mind is the mtz labels: I(+), sigI(+), I(-), sigI(-) merged. What does 
>> > merged mean? Could this be the reason? Other issues that could causing 
>> > the
>> trouble?
>> The "merged" means that the input data were only partially merged -
>> this usually happens when processing in HKL2000 using the "no merge
>> original index" setting, where the reflections are not merged to the
>> asymmetric unit (anomalous or not).  I've never used XDS, but I guess
>> it must do something similar.  Phenix doesn't really deal with data
>> like that; it always merges equivalents (while leaving Friedel pairs
>> alone by default).  This usually doesn't have any impact on the
>> anomalous signal.  If XSCALE has an option to merge the data more
>> completely, this should make the "merged" tag go away.
>I went through the xds menu again and found that the "merged" seems ok in
>terms of anomalous signal.
>> It sounds like your element of interest isn't very well ordered; do
>> you see the rest of the ligand in the normal maps?
>I was hoping to identify the ligand binding position by resolving the
>anomalous map first. Shouldn't this be the procedure to locate ligand?
> If you really want
>> to be sure that it's not a data-handling issue, you could try
>> reprocessing in other programs and confirming the result.
> I know CCP4 can also generate anomalous difference map. But I myself have 
> never done it (I googled online and found it not that straight forward). 
> Can anybody go through it for me please, or there are other handy 
> programs that can make anomalous difference map?
>>  I don't
>> know if radiation damage could be at fault, but it's always a
>> possibility.
>> -Nat
>Structural Biology Department
>University of Pittsburgh

More information about the phenixbb mailing list