Chemical name annotators that find and standardize chemical names in scientific papers and patent text documents usually generate tables of SMILES or InChIs. In order to make the most use of identified compounds for text analytics, we convert the found compound names into InChIKeys and insert the InChIKey values into the original document text beside the compound name, along with the literal word 'inchikey' as an entity type marker.  The augmented documents are then indexed as full text using Solr/Lucene resulting in a search service with useful capabilities such as co-occurence analysis, e.g., finding all cases of any text synonym of aspirin within 10 words of cancer, finding cases of any chemical compound (any entity marker 'inchikey') within 10 words of 'produced' AND within 10 words of 'yield'.  This indexing technique also supports finding chemically similar documents.  The selected compounds' InChIKeys are entered together as a set of query 'words', returning an engine-ranked list of documents with the most compounds overlapping the target's. Full engine capabilities such as boolean filter terms are still available to refine the set with full engine performance.  The indexed augmented text also facilitates other data mining capabilities such as batch searches for large numbers of molecules over millions of documents.Our team is developing a comprehensive chemical pedia containing structure representations and attributes for millions of molecules derived from patents and other sources such as Medline Abstracts. On-going work includes entity insertions simultaneously performed on gene, drug and disease annotation types, enabling rich entity/text combination queries with no search engine modifications needed.

InChIKeys as chemical entity ids to enable in-context text indexing and to identify engine-ranked chemically similar documents

248th National Meeting (2014)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.