I just scrapped most of my answer, Robbie was quicker, and I guess
Gerard is on holiday ;-)
As 'honorary Dutch' though here are my two cents.
It is only fair that a well-informed and well-educated human
being can do a better job than a fixed-frozen automated procedure.
This is exactly the whole problem Niko.
I must say that a certain proportion of the PDB depositors, are
neither well-informed nor well-educated.
Else, I cannot see why anyone in his right mind would put 'waters'
1.8 A from a Ca in 2009 (see Robbie's email) or
create O-O clashes of 2.4 A in 2008 (where a simple peptide flip would
make two perfect hydrogen bonds), etc etc.
Would be easy to blame it to students that spend three or four years
cloning before they get a structure
that ends up with many mistakes in the PDB. PIs here carry most (if
not all) of the blame: many are simply not
competent to supervise crystallographic projects (see some CCP4
questions to see what I mean ...),
a few are too busy (thus they should not be supervising PhD students
if they lack the time), some make
honest mistakes (ehm ... just in case somebody finds any errors in my
lab's structures ...), and many do a
good job and educate good students. But errors in the PDB accumulate.
Recently, I had a look, together with Robbie, at my good-old chitinase
(1994). Thats a pre-likelihood structure.
Simple re-refinement with Robbie's tools (and some new tricks, also
automated) and the script does in this 2.3 A
structure a better job than I did as a PhD student, pushing it from
the 48% of MolProbity to 98%.
And I dont think that me or my supervisors could had done a better job
at that time,
sharing night shifts in an Indy and an ESV, with PROLSQ/PROTIN. So, I
am more than happy
if an automated procedure in 2009 can tell me that this job can be
done better with new tools.
Finally, validation tools are not there to pass judgement. They are
tools to be used by depositors, referees, and users
alike, to help them make a better informed interpretation of
crystallographic *models*. Servers like EDS and PDB_REDO
must be seen in that light,
Take care -
A.
