How can I do a quick performance check of how fast and reliably my cluster can generate a large Phase database?

By default the conformer generation is done in blocks of 5000 entries. Therefore if you have N CPUs available (and licenses to run N phase jobs), then create a test input file of N x 5000 compounds. This can be a smiles, SD or maestro file. Now generate the Phase Database using the same workflow which you intend to use for the full database generation. This can be the "Generate Phase Database" workflow from Maestro via Applications → Phase or the backend commands phasedb_manage, phasedb_confsites or a script of these and more commands, like para_ligprep or filtering commands like ligparse or a KNIME workflow.

It is important that you are using the same workflow that you intend to use for the large database generation. This allows you to identify bottlenecks of your workflow, check if your cluster or network might have problems and check the scientific quality by running pharmacophore searches on this test database. Also you have a better overview of where your workflow might be improved for speed or accuracy (filtering, number and quality of conformers, for example).

Keywords: phase database, phasedb_manage, phasedb_confsites CPU compounds

Back to Search Results