On 15/09/11 17:47, Engin Özkan wrote:
Dear coot users and developers,
We are using coot 0.7-pre-1 (rev 3633) in the lab on linux and Mac
machines, and have observed something odd that we did not see before.
Most (all?) pdb files when opened give a (harmless) warning that says
"Molecule x has nomenclature errors. Correct them?" The residues
mentioned are Phe's, Tyr's, and Asp's. I don't think there was an update
to these residues in the latest dictionary. Before I go into more
detail, is anybody else seeing this, and should we just install a newer
nightly?
I should note that these models have been refined by Phenix versions 1.6
and 1.7, and molprobity thinks they are perfectly fine. There are no
DNA/RNA bases, and odd ligands, except sugars in them.
I was away for the day.
Back now.
I realise that this can be somewhat of a surprise. The problem was that
when setting torsion angles for rotamers (and referenced the "labeled"
atoms that Dale mentioned) we need to make a choice of atom - when there
are nomenclature errors this can leads to completely incorrect side
chain-positions. The options were
1) silently incorrectly set some torsion angles [a]
2) use a dictionary that enforced chiral centres for LEUs and VALs
(unlike the standard Refmac dictionary) [b]
3) silently "correct" the nomenclature errors [c]
4) warn the user of nomenclature errors - and provide an easy means of
reseting them.
[a] this is the way that it used to work - and was though inappropriate
by several experienced users
[b] this was an attempted solution and seriously disapproved of by my
CCP4 colleagues
[c] I didn't like this - Coot shouldn't secretly and surprisingly move
your atoms. And as Clemens says, it is not always correct to reset the
nomenclature "errors".
I believe (hope?) that I chose the least worse option.
Paul.
