On Wed, 2007-05-23 at 15:58 -0400, Jianghai Zhu wrote:
> Hi all,
> The ramachandran plot in Coot is a very nice tool, especially it is
> based on Richardson's 500 database. However, I just compared the
> ramachandran plot in Coot and the ramachandran plot generated by
> RAMPAGE and MOLPROBITY and I found out that they are actually slightly
> different. A few residues which are outliers in Coot are actually in
> the allowed region in the ramachandran plot generated by RAMPAGE and
> MOLPROBITY and vice versa.
I suspect a quantisation issue.
The contour levels in Coot are 2% and 0.12%. The numbers are pulled
from a clipper class:
http://www.ysbl.york.ac.uk/~cowtan/clipper/doc/classclipper_1_1Ramachandran.html
And FWIW, I also like my variable names to line up.
> Also the contours of the allowed region of different categories of
> amino acids look a bit different in Coot. I noticed that the
> ramachandran plot in Coot is not smooth. This might contribute some.
> But why is there a hole in the Gly category coutour around (Psi, Phi:
> -70, 110)?
Isn't that just a contour level issue again?
> and why the contours of Pro and general residues are actually
> different from the ramachandran plot published in the paper (S.C.
> Lovell, etc, Proteins: Structure, Function and Genetics 50, 437-450)?
AFAICS Lovell et al. contour at 2% and 0.2%. The 2% level does look a
bit different, I admit. I wonder if that is because they exclude
pre-pro (which excludes some pros) and those in clipper do not. I think
Lovell et al. also adjust their sampling around sharp edges.
Frank von Delft was querying this a little while ago:
http://www.ysbl.york.ac.uk/~emsley/coot/mbox-2004-2005/0201.html
Seeing as Rampage also uses 2% and 0.2% (AFAICS), I feel an urge to use
those levels too. I welcome feedback about this.
Paul.
|