Filtering based on graph (and possibly physical) properties should always be carried out prior to generating multi-conformer databases using OMEGA. Eliminating undesirable compounds prior to generating conformers will save execution time of both OMEGA and downstream applications, and space on disk. Large polypeptides or proteins, very flexible molecules, or simply molecules that would never be considered useful for the ultimate modeling application are best eliminated from a data set as early as possible.
The most important graph filter to apply is rotatable bond count. Although OMEGA may be able to generate conformers for molecules with more than 20 rotatable bonds, the results of such an exercise would be dubious at best for any conformer generation method. Number of rings, especially flexible rings, should also be used to exclude irrelevant molecules. OMEGA will handle molecules that have many thousands of possible ring combinations, although the time expenditure may be prohibitive and the results equally questionable to molecules with an unreasonably large number of rotatable bonds. Simple element filters may be useful as well, although OMEGA will discard compounds for which no force field parameters exist. Using element filters beforehand may simply aid in tracking the rationale for discarding compounds instead of searching through OMEGA log files for failure modes.
The FILTER program from OpenEye, Cadence Molecular Sciences provides all of the functionality outlined above, and additional physical property filters. It is highly recommended that FILTER or a program similar in functionality be used for input file preparation of large datasets.