Well - it isnt surprising that all your geometry is "good" at the start.
You have fitted a refined structure against a a different crystal form,
so the first geometry report relates to your starting model which will
not be the true model which fits your new data.
Refinement has to push that model about to get to the truth, and
inevitably that will degrade the geometry. That is good - not bad..
My usual approach is to let the refinement rip as long as the Rfactors
are falling, then use the "bad" geometry indicators to pinpoint where
things need to be changed. You will find many such regions where side
chains have moved, and even loops - often very easy to correct in coot.
If you cant see what to do set the occupancies of the changed reegions
to 0.00, do a few more cycles of refinement and see if the difference
maps indicate where that part of the structure should go..
Eleanor
On 06/22/2011 07:25 AM, Careina Edgooms wrote:
Dear ccp4 members
I have something that surprises me with molecular replacement. I have obtained a
solution for a single point mutation using Phaser, the solution seems ok. I do
one round of refinement with refmac and I check the structure using molprobity
before I even start really to refine it. The structure looks very good.
MolProbity gives all green lights. Then I start to "fix" the structure, I add
waters, I check fit to density, rotamers, geometry etc, I do some more
refinements. I check with MolProbity and it looks a lot worse... many clashes,
many bad ramachandran and rotamers, many red lights. I do not understand. How
can each successive round of refinement make the structure worse and worst? Is
there a fundamental problem, perhaps, like undiagnosed NCS or incorrect space
group or incorrect MR solution? Could they be giving these strange result do you
think?
Thanks
Careina
