Marcus,
it appears that coot "breaks" bonds when the distance between atoms
exceeds 1.64A pretty much irrespective of the bond type. Two exceptions
are, of course, sulfurs in MET/CYS. CB-SG bond in cysteine does not
seem to have any cutoff (when you do rotate/translate zone in coot, you
can move individual atoms around while holding ctrl, and at least in my
hands this bond is never broken, up to 150A which is infinity for all
practical purposes). SD-CE and CG-SD break when stretched to 3A, and
hydrogens are allowed to wander away to 1.42A before the bonds are
"broken".
It is of course more complex than a simple calculation, but if we simply
assume that bond lengths vary according to normal distribution centered
around their target values and esd's listed in monomer libraries, then
ten most "vulnerable" bonds are
PRO CB-CG 1.492(0.050)
ILE CG1-CD1 1.513(0.039)
THR CB-CG2 1.521(0.033)
VAL CB-CG1 1.521(0.033)
VAL CB-CG2 1.521(0.033)
ILE CB-CG2 1.521(0.033)
LEU CG-CD2 1.521(0.033)
LEU CG-CD1 1.521(0.033)
ALA CA-CB 1.521(0.033)
THR CA-CB 1.540(0.027)
While some of the "usual suspects" are in the top ten (in my experience,
most often broken bonds are in ILE, THR and VAL), I have frankly never
seen a broken proline after refinement - which probably means that
electron density for these is genuinely better than for oft-disordered
isoleucines. Same goes for Ca-Cb bond in alanine.
Guess my point is that broken bonds in coot are not necessarily much
worse than those that are not. Think of it this way - it only takes
about 3.25 sigma deviation for the ILE CG1-CD1 bond to "break", whereas
the CG-OD bond in aspartate may be stretched ~20 sigmas from its
expected length and still appear intact.
Cheers,
Ed.
On Mon, 2011-01-17 at 10:28 +0100, Marcus Fislage wrote:
> Dear all,
>
> I might excuse myself for the silly question but it is the first time I
> solve an x-ray structure.
>
> After modell building in coot and running of refmac with restrained
> refinement I have the problem that the pdb output file contains
> distances between e.g. ILE Cb and Cg that are so long that the Cg is
> displayed as single atom (distance ~1.64 A instead of 1.5 A). This seems
> to happen to me for aminoacid sidechains where I added an alternative
> conformation (especially if the 2nd conformation is oriented not far
> away from the first one) and if I set the occupancy of sidechains like
> Glu for Cd and further to zero (Here refmac gives a bond break between
> the last atom with occupancy 0 and the first with occupancy 1). I can
> adjust it of course in coot back to normal but after the next refmac run
> the same happens again. The option card in refmac for geometric
> restraints I kept untouched.
> Am I missing to set on an option in refmac, or is there something else
> going wrong?
>
> Thanks a lot for the help
>
> Marcus
--
"I'd jump in myself, if I weren't so good at whistling."
Julian, King of Lemurs
