There are many benefits to being on the CHMINF and CCL distribution lists. I recommend them. At the end of today an announcement regarding OSRA dropped into both of them. OSRA – Optical Structure Recognition Application.
It is declared as: “a freely available software is a new addition to the set of chemoinformatics tools available from the Computer-Aided Drug Design Group at the NCI-Frederick. OSRA is a utility designed to convert graphical representations of chemical structures, such as they appear in journal articles, patent documents, textbooks, trade magazines etc., into SMILES. OSRA can read a document in any of the over 90 graphical formats parseable by ImageMagick (GIF, JPEG, PNG, TIFF, PDF, PS etc.) and generate the SMILES representation of the molecular structure images encountered within that document.
Rzepa et al have reported previously on “machine vision methods for the identification of chemical composition diagrams from two-dimensional digital raster images.” It’s a good foundation document for understanding some of the challenges with converting structure images to a chemical structure format
I’ve had a lot of interest in these technologies in relation to other work I’ve participate in – the encoding of chemical structures into barcodes (Patent Pending 10/382,461) with the intention of putting chemical databases literally in your hands.
With all these efforts, and based on my exposure to these technologies over the years, my comments are…it’s definitely work worth doing and the efforts have only been partially successful at best (again based on my personal experiences). So, with the announcement of OSRA, and with much excitement I started hitting the online system with challenges.
My observations to date applied to exports of 300dpi TIFFs from Chemsketch
1) Hexylbenzene converted perfectly..as did many other simple structures…simple chains and rings.
2) Bridgehead compounds are problematic…crossed bonds are the issue it appears
3) Stereobonds cause problems…maybe at conversion…maybe with SMILES handling (to be investigated!) Less than 50% success at retaining appropriate stereobonds…while quick visual inspection suggested structures were the same only alignment of the molecules and careful examination showed that in a number of cases they were not matching. The easiest was to compare was comparison of systematic names or InChI.
4) Organometallics were problematic…an example shown below. The pair of structures on the left were derived from the conversion of the structure on the right. Notice the Al was converted to Ar (aryl?), a bond was severed, an oxygen atom was lost.
OSRA is clearly a work in progress…but a valid and worthy effort I will support with great enthusiasm. From the point of view of ChemSpider I imagine a time when OSRA is ripping it’s way through PDF files and converting structures on the fly, with a very low error rate and extracting those structures from Open Access chemistry articles to populate ChemSpider and make those Open Access Articles searchable by both structure/substructure and of course by text through the ChemRefer search engine.
That being said…what makes all of this easier is for Adobe to collaborate with the chemistry community and embed InChI tags or even connection tables into the PDF format. A number of years ago ACD/ChemSketch was extended to allow the export of PDF format reports with the structure connection tables tagged in the PDF format. All ChemSketch PDF generated files are therefore structure searchable. Anybody out there have connections into Adobe Life Sciences?
ACD/Labs have just done the same for PNG files by embedding the InChI codes directly into the PNG format when exported from ACD/ChemSketch. This was at the request of the Wikipedia community.
So…my commendations to Marc Nicklaus and Igor Filippov at the Computer-Aided Drug Design Group at the NCI-Frederick. OSRA release 1 looks like it is heading in the right direction but has some hills to climb yet. SIgn me up to help.