Title:

Enhancing OMEGA With Experimental Data From the CSD

Authors:

Jens Sadowski (a)
Jonas Boström (b)
(a) GSI Chemical Computing,
(b) Department of Medicinal Chemistry, 
AstraZeneca R&D Mölndal, Sweden

Abstract:

The conformer generator OMEGA is widely used as an efficient
and fast way to generate multiple conformations of drug-like
molecules. OMEGA searches torsion angle space by assigning torsion
angles to rotatable bonds via SMARTS patterns. The default torsion
data file is a handcrafted set of 123 SMARTS patterns.

Here, we introduce an automatic procedure to obtain many more SMARTS
patterns (>3500) from experimental structures in the Cambridge
Structural Database (CSD). This enhancement reduces significantly
both the number of missing torsion patterns (from 30k to 1k)
and the average RMS (from 0.30Å  to 0.21Å and
increases the coverage of reproduced X-ray structures from 83%
to 93% over our dataset of 24k X-ray structures.