You know, I've been pondering that question for most of my adult life.
Why can't we push macromolecular R factors down to the level of
experimental error like our "small molecule" colleagues do routinely? I
have a few ideas, and others do too. In fact, there will be session on
this topic at the next ACA meeting.
But right at this moment the best advice I can give you is to load up
your model and map in coot and use the Validate : Difference Map Peaks
option. The highest and lowest points in your FoFc map are the least
likely of all difference features to be due to noise. Look at them, and
try to do something about them. When you run out of ideas, you're
done. Then you look at your statistics. If the R factors are "too
high" and the geometry is way too good, then you need to increase the
x-ray weight in refinement. Once you've changed something in the
refinement options, let it run for a while so the programs can do their
job. When things stop moving it is time to look at maps again. The
myriad of options in refinement and automatic building features are all
things that can save you time in this process, but at the end of the
day, it is the difference map that tells you what to do next.
-James Holton
MAD Scientist
On 9/15/2017 4:32 AM, rohit kumar wrote:
> why R/Rfree not going down from 21/25?
|