Anomalous map average by NCS

From: Hui Wang
Date: 22 September 2011 00:45


Dear all,

I am wondering if anybody did NCS average for Anomalous map. Based on
my knowledge, there are two days to do it.

1, Use Map average in CCP4 program (input from map which cover unit
cell, create averaged density for whole unite cell )
2, Use NCS map in coot program.

However, I got different results from both ways. I am not sure which
one is corret. Someone can help me with that?


Thanks  a lot

Phoenix

----------
From: Jens Kaiser


Hui/Phoenix,
 It is pretty hard to say from what you describe which method is right.
The NCS maps in coot will give you a correct - but not as good of an
estimate of your density - if, and only if, you have enough of a model
and you have defined your NCS master and "slave" chains correctly. If
you have chemically different but similar subunits and you don't check
them correctly in the NCS ghosts dialog, you'll get bad results.
 CCP4 map average (from the map and mask utilities) is pretty tricky,
most of the times;
 When I had to do averaging of anomalous difference maps, I found it
most useful to do cyclic averaging of your "normal" (i.e. non-anomalous)
map, either with dm, ave/rave or MAIN (and probably phenix, though I
have not tried it); that will give you good indications of the quality
and validity of your averaging (i.e. backtransformation R factors;
correlation coefficients tend to be also more reliable). Then use these
phases (rotated, but the ccp4i fft task will do that for you if you
select "anomalous map") to calculate your "averaged anomalous map" with
your DANOs and the averaged phases.
 This seems to me the most robust and reliable method of averaging
anomalous maps in the general case. If you have your NCS related
molecules defined properly, you might want to add coot's average maps on
top of that, but it is unlikely to improve the outcome.

Cheers,

Jens

----------
From: Eleanor Dodson


The cooot method is easiest, as long as you have model coordinates. Then coot works out the transformation between molecule 1, 2, etc  and averages any map requested to cover molecule 1.

However if you dont have a model it is very tricky to get a proper transformation matrix.
If you have enough heavy/anomalous sites programs like profess will try to suggest transformation relating one subset to another.

If you have some phases then buccaneer will try to build a model and look for any acceptable NCS transformation.

Once you have the transformations mapaverage will work..
Eleanor