Proceedings,
Improvements in Optical Structure Recognition Application
(2010)
Abstract
We present recent improvements of the Optical Structure Recognition Application (OSRA), an open source utility to convert images of chemical structures to connection table type description in an established computerized molecular format. There exists a large body of chemical information which has remained largely inaccessible to machine data mining techniques so far. One of the most common ways of describing a chemical structure in a journal publication or a patent document is by drawing a two-dimensional structure diagram which represents atoms and bonds of the molecule in a human-recognizable form. While easily interpreted by a human expert, such drawings are by themselves unsuit- able for use in a computer database for applications such as virtual screening and computer aided drug development. OSRA allows recognition and conversion of such drawings into computer formats widely used by the chemoinformatics community. This paper describes recent progress we have achieved for OSRA in terms of faster processing times and more accurate recognition rates.
