Michael Thompson wrote:
Hello All,
I am currently solving a structure at 2A resolution with phases obtained from molecular replacement. Using the MR solution, I began refinement with Refmac using NCS restraints. I am currently building the parts of the model that were left out of the MR search model, and have just about successfully completed all three NCS-related chains. Obviously I will continue to use the most complete model for refinement, and plan to release the NCS restraints over the parts of the molecule that don't quite seem to obey perfectly.
My question is, once I have connected density for all three chains will it
still be worthwhile to perform density modification, such as solvent
flipping/flattening or histogram matching (implemented through SOLOMON and/or
DM) to improve phases? It seems that I have always been told that density
modification is typically carried out at the beginning of refinement, prior to
any model building, however my understanding of these types of density
modification (particularly solvent flipping/flattening) leads me to believe
that they would be most effective when more of the molecular envelope can be
identified, such as during later stages of refinement.
My feeling is that density modification is most useful when the model is
substantially incomplete (or substantially wrong). But I'd use both the
unmodified and modified map when trying to interpret the difficult bits.
'dm' is obsolete, I'd use 'parrot' by preference.
Also, I read something recently that lead me to believe that the solvent
flattening procedure may be implicit in the implementation of NCS averaging in
refinement software. I understand that the two processes are fundamentally
different and independent of one another, but the information I recently read
described something like the following (unless I misinterpreted). Because real
space NCS averaging requires identification of the molecular envelope in the
same fashion as solvent flattening, during NCS averaging the envelope is
identified then the map is solvent-flattened and averaged using the NCS
operator. I am unfamiliar with the inner-workings of most crystallographic
software, so I was wondering if this is how NCS averaging is implemented in
Refmac? I suppose another way to ask the question would be: If I have an
NCS-averaged map from Refmac, is it already solvent-flattened?
Refinement with NCS restraints essentially imposes the NCS in a way
which is different but redundant with density modification, so I
wouldn't expect further density modification after refinement with NCS
to help much. The same is somewhat true for solvent flattening too.
The only thing density modification might get you at this stage is break
you away from model bias a little. Again, it's so quick you may as well
try it and look at both maps.
Kevin
--
EMAIL DISCLAIMER http://www.york.ac.uk/docs/disclaimer/email.htm
