Title:

Smidgen: A simple program for the production of dictionaries for
crystallographic refinement using OMEGA and OEChem.


Author:
Joel Bard
Wyeth Research

Abstract:

The production of x-ray crystal structures requires iterative refinement
of atomic coordinates against diffraction data.  At the diffraction
resolutions usually encountered in macromolecular structure determination,
the observation to parameter ratio is insufficient to constrain the atomic
model such that well established bond lengths and torsion angles are
maintained through the refinement process.  To compensate for this,
crystallographic refinement packages include empirical geometric restraint
terms in their energy calculations.  The parameters for these geometric
restraints are standardized for all of the amino acid and nucleoside
residues encountered in biological macromolecules but need to be generated
de novo for any non-standard compounds which may be bound to these
structures.  The generation of these restraints is obviously of particular
importance in pharmaceutical crystallography since geometric restraints
must be generated for the compounds in any structures of drugs in complex
with their targets.  Currently, publicly available programs to
automatically generate restraints dictionaries for small molecules use PDB
files as their input.  This causes problems since these pdb files rarely
specify connectivity and cannot specify bond order.  It is therefore
impossible for these programs to unequivocally determine which bonds are
rotatable, planar etc.  In the developmental package AFITT, Openeye has
implemented dictionary generation starting from a SMILES string.  I have
independently written smidgen, a similar scheme which is run from the
command line.  Given an identifier from our corporate database, smidgen
retrieves the SMILES, adjusts the protonation state to what would be
expected at neutral pH and then runs OMEGA and Szybki to generate a low
energy conformer.  The core of the program then generates dictionaries of
restraints for bond lengths, bond angles, torsion angles, chirality, and
planarity for the two most popular refinement packages, refmac and CNX.
PDB files are also produced for importation into molecular graphics and
ligand fitting packages.