You havent got an incorrect CISPEP definition lurking in the PDB file
have you - COOT does NOT correct these so if you rebuild something which
has been previously flagges as CISPEP ( and that can happen if a very
ugly omega angle falls to 89.999 ) then it stays that way in the PDB
output from COOT, and REFMAC struggles valiantly to reconstruct a CISPEP..
Eleanor
Jan Abendroth wrote:
Hi all,
along the lines of a recent discussion:
At the final stages of the refinement of a structure which a whole bunch of
ncs (2x4chains) at 2.2A resolution in P1, I run into the following very
annoying and persistent problem. The Ramachandran plot shows a whole bunch
of ugly outliers. While some of them appear to be "true", ie. the same
outliers in all chains w/o application of ncs, loop region, H-bonds that
make sense, ..., there are a number of "misbehaving" outliers: After
reciprocal space refinement some residues violate the Ramachandran plot for
one or two chains only. Both 2FoFc and FoFc density clearly show that the
main chain is supposed to be at a slightly different position. Real space
refinement in coot puts the mainchain nicely in place on top of the ncs
mates' position. Refmac then pulls things back into the forbidden region.
For an illustration see here:
http://picasaweb.google.com/Jan.Abendroth/RefmacAndRamachandran/
I have tried to tie down the violators with tight ncs restrains - no
success. I am running refmac 5.3.0037, with TLS refinement. Rfactors seem
really decent (20%/25%) as does geometry. Refmac complains about
"MAKE_U_POSITIVE" problems.
Any ideas?
Cheers
Jan
