[phenixbb] hydrogens in riding model create steric clashes after refinement

[Nathaniel Echols]

On Thu, Aug 19, 2010 at 3:46 PM,  <amadrona at uci.edu> wrote:
>No, I didn't previously have hydgrogens.  This is the first time I looked
> at close contacts with hydrogens present.  I submitted my pdb to the the
> PDB auto Deposition input tool which defines "close contacts" as less
> then 2.2 A
> After looking at these I could see that many were between hydroens of one
> side chain and usually an amid or sometimes CA of the mainchain.
> Sometimes they appeard hydrogen bonded but the hydrogens had to be rotated
> away.  All I had to do was rotate the hydrogens.  I was just suprised that
> this would not have been done in automatically in the refinement.

If you didn't have explicit hydrogens, there's no way for
phenix.refine (or any other refinement program) to tell that there's a
clash.  Minimization algorithms are "dumb", in the sense that they
know nothing about chemistry, only restraints, and rotating around a
bond often requires passing over large energy barriers (a major reason
why simulated annealing is used).  Recent versions of phenix.refine
can optionally re-fit sidechain rotamers and flip N/Q/H residues to
minimize bad contacts, and more features like this will be added over
time, but it's tricky.

It's very suspicious that hydrogens are bumping into the C-alpha if
there is space for them elsewhere; the tools in Phenix should almost
never do this unless there is something else wrong with the model.
But it's difficult to guess without seeing the structure.

-Nat