***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***


Hi Nick,

> in a structure that i am currently working on via mol rep, i see
difference density for a peptide ligand.  when i build the ligand and do
refinement in REFMAC 5 and look at the 2fo-fc map, I see nice density for
> the ligand now.  but, if I then use the coordinates with the ligand to
look at a density modified map within CNS, i only see density for bits of
> the ligand.  does this mean that the 2fo-fc map is wrong or that the

This turned out longer that I expected...the short version is to check the
mask used in cns solvent flipping (may be assigning ligand region as
solvent), try pirate, and read B.C. Wang's Methods in Enzymology paper.

Longer version follows...

Regarding density modification - are you using model phases or
experimental phases as your original phase set?  I'm personally cautious
regarding density modification of model phases, although it seems to be a
reasonably common practice.

Another thing to keep in mind is that there are several different density
modification schemes, and they can behave slightly differently.  If I
remember correctly, cns uses solvent flipping (solvent flattening with
gamma correction, similar to dm).  Your ligand density may be getting
classified as solvent during mask generation (if you're generating your
mask by spherical averaging, then this is a possibility): you could try
changing a range of different sphere radii for mask generation and see how
the map changes.  If you are generating the mask from the protein model,
then the ligand might be getting cutoff there as well (if this is the
case, I don't have any suggestions for how to proceed without introducing
bias into the maps...although if you're using a model-derived mask on
model phases, I'd be skeptical that it would be of much use).

With input experimental phases, manually setting part of the protein
region to solvent weakens the density but doesn't eliminate it, so your
suspicion regarding ligand disorder could be reasonable.  I haven't tested
this with model phases, so I don't know how this would behave.

You could try pirate, which uses a density modification procedure that
doesn't use a mask (my simplified understanding is that it's essentially
maximum likelihood histogram matching, but this is probably an
oversimplification).  If you do try this, I'd recommend testing several
different reference structures.  When I was testing it, the solvent
percentage of the reference structure seemed to have the largest effect on
the final map, but this may vary with your data.

As far as references go, B.C. Wang's Methods in Enzymology is a very clear
introduction to solvent flattening, although it doesn't cover more recent
developments (gamma-correction, histogram matching, etc).

Good luck,

Pete

Pete Meyer
Fu Lab
BMCB grad student
Cornell University






Reply via email to