*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
Dear All,
This thread has already grown too much, so I should start by apologising for
adding yet another long message to your mbox.
Assume, for the sake of a thought experiment, that instead of the safe
academic environment of my office, I find myself with gun pointed straight
into my head and a European grant committee (across the room) demanding from
me to re-determine my latest structure. What they want me to do, is to
produce a new PDB file that will represent in the best possible way _not_
just the experimentally determined electron density map, but everything that
I know about the problem in hand (hint: what have I deposited ?). Assume,
further, that they explicitly demanded that I only use computational methods
(no more experiments, sorry). What should I do with those 'missing'
surface-exposed highly mobile side-chains ? I would definitely not declare
them 'missing' (unless, of course, I knew better based on chemical,
biological or other insight). Similarly, I would not just use the most
highly populated rotamer from a library, mostly because such a treatment
completely ignores the different environments of the specific side chains in
the given structure. Because they asked me for a PDB file, re-inventing a
new structure-description format is not an option. What should I do (for the
life of me)? Having thought about it for a while [a gun pointing straight
into ones' head seems to help clarifying ones' thoughts; we should possibly
try this with doctoral students], I believe that I would have ended-up doing
molecular dynamics: Take the structure, add all missing side-chains in a
sensible (but otherwise arbitrary) conformation and do a couple of long,
state-of-the-art molecular dynamics simulations (state-of-the-art meaning
with a modern force field, full electrostatics, explicit waters, etc). The
dynamics of individual side-chains (assuming that the core of the structure
remains stable) are well-within the time scale of todays' molecular dynamics
simulations (and you can always confirm convergence via a principal
component analysis). Then, align the missing side-chains on the backbone and
do a cluster analysis of their MD-derived conformations to identify the
major conformers. The relative residence times (for each conformation) would
be deposited as occupancies, the atomic fluctuations would give the
corresponding temperature factors. And I would then have to hold my breath
waiting to see what the committee thinks about this chimeric (experimental
plus theoretical) model of mine.
Again apologies for the rant,
Nicholas
ps. Hard-core crystallographers ready to dismiss molecular dynamics as
"fantasy, pure-and-simple" will be reminded that a significant portion of
PDB (ie. NMR structures) is the result of the restrained application of
molecular dynamics force fields (and that, of course, molecular dynamics is
indeed fantasy, pure-and-simple ;-).
--
Dr Nicholas M. Glykos, Department of Molecular
Biology and Genetics, Democritus University of Thrace,
Dimitras 19, 68100 Alexandroupolis, GREECE, Fax ++302551030613
Tel ++302551030620 (77620), http://www.mbg.duth.gr/~glykos/
