By Stephen Boyer (IBM Research)

As a part of our long term project on computer curation of patents and the scientific literature, we have aggregated a substantial repository of millions of chemical compounds. We successfully employed ChemAxon tools build a very fast similarity search and substructure / substructure search engine for searching these molecules using dB2 (9.5) and JChem Base (5.3). We started with hybrid SMILES strings derived from a variety of sources including – Daylite, Pipline Pilot & CambridgeSoft. These SMILES were processed using ChemAxon SW into CA – JMoL Zip files and added to an all molecules chemical repository. The JMOL Zip files were further used to generate chemical fingerprints (16 bit numbers). Both the JMOL Zip files and the chemical fingerprints were stored in our dB2 tables – and are searched using web services that access J Chem Base via a Java API to perform the searches and integrate the results into our web based application (currently called SIMPLE). We will demonstrate this application to access > 10 M unique compounds – contained in > 18 Million documents.

Click here to view the complete document or click here to view the video.