Lowering the matrix term further might help. I've set it to 10
when working with ~3 A RNA maps, so you might want to try values around
5-10. If you're using the most recent build of Coot (revision 3440),
you could also try using harmonic restraints. These allow an atom to
move slightly during real-space refinement, but try to keep it close to
its starting location.
Python: add_extra_start_pos_restraint(1, "A", 150, "", " CA ", "",
0.25)
Scheme: (add-extra-start-pos-restraint 1 "A" 150 "" " CA " "" 0.25)
will add a start-position restraint to molecule 1, chain A, residue 150,
no residue insertion code, alpha carbon, no atom alt conf, 0.25 Angstom
standard deviation. For standard deviation, the higher the number, the
more the atom is allowed to move. Try 0.25 or 0.1 for fairly strong
restraints. There's currently no easy way to add harmonic restraints
for all atoms at once, but you could try adding restraints for all alpha
carbons in a trouble region and see if that helps. If you want to
remove a restraint you've added, use
delete_extra_restraint(1, ["start-pos", ["A", 150, "", " CA ", ""]])
(delete-extra-restraint 1 (list 'start-pos (list "A" 150 "" " CA
" "")))
to remove one at a time or
delete_all_extra_restraints()
(delete-all-extra-restraints)
to remove them all.
Hope that helps,
Kevin
On 3/18/2011 3:53 PM, Florian Schmitzberger wrote:
Hi everybody,
Have there been any advances or are there specific settings in COOT
for low-resolution regularization or real-space refinement, along the
lines of lego in the O program? I have found some old threads on the
topic, such as this one:
http://www.ysbl.york.ac.uk/~emsley/coot/mbox-2006/0640.html
I am having trouble using COOT, building into 3.8-4 A maps (the maps
are rather good for the resolution; 2-fold NCS, high solvent content),
for the idealization of beta-strands in B-factor sharpened maps. This
does not work well in COOT for me, the mainchain is frequently moves
outside of the density. I did reduce the matrix term to 20, but keep
having trouble.
Any suggestions are very welcome.
Cheers,
Florian
