{"id":1852,"date":"2016-07-28T13:21:24","date_gmt":"2016-07-28T12:21:24","guid":{"rendered":"https:\/\/nextmovesoftware.com\/blog\/?p=1852"},"modified":"2016-07-28T14:05:42","modified_gmt":"2016-07-28T13:05:42","slug":"sketchy-sketches","status":"publish","type":"post","link":"https:\/\/nextmovesoftware.com\/blog\/2016\/07\/28\/sketchy-sketches\/","title":{"rendered":"Sketchy Sketches"},"content":{"rendered":"

Chemical structure diagrams are essential in describing and conveying chemistry. Extracting chemistry from documents using text-mining (see NextMove Software’s LeadMine<\/a>) is extremely useful but will miss anything described only by an image.<\/p>\n

As a general approach to mining chemistry from images, one may consider using image-to-structure programs such as: OSRA<\/a>, CliDE<\/a>, ChemOCR<\/a>, and Imago OCR<\/a>. However, image-to-structure is not easy or quick and can be prone to compounding errors (e.g. OCR<\/a>).<\/p>\n

At NextMove we approach this problem slightly differently. It turns out that in some cases the source sketch files used to create the chemical diagrams may be available and provide a ‘cleaner’<\/i> data source than the raster images.<\/p>\n

Although the data is ‘cleaner’<\/i> in terms of digital representation, na\u00efvely exporting the connection table stored in a sketch file can lead to artificial and erroneous structures. The main problems stem from the stored representation (connection table) imprecisely reflecting what is displayed. To account for these issues, the NextMove Software<\/b> converter (code name: Praline<\/b>) applies correction, interpretation, and categorisation to sketches. The transformed connection table (currently written as ChemAxon Extended SMILES<\/b> [CXSMILES]) better reflects what is actually displayed.<\/p>\n

Let’s take a look at what’s possible with three examples:<\/p>\n

1) US 2015 344500 A1<\/b><\/p>\n

Method 9 in US 2015 344500 A1<\/strong><\/a> describes a four step synthesis:<\/p>\n

\"US20150344500A1-20151203-C00112\"<\/div>\n

Using image-to-structure SureChEMBL extracts four structures, I’ve added the titles to make it easier to pair up:<\/p>\n

\"SCHEMBL17309138\"
\nCompound 2-2 (OCR error)
\nSCHEMBL17309138<\/a> \/ CID 118554493<\/a>
\n\"SCHEMBL12363\"
\nCompound 9-1 (part)
\nSCHEMBL12363<\/a> \/ CID 10008<\/a>
\n\"SCHEMBL17307813\"
\nCompound 9-2
\nSCHEMBL17307813<\/a> \/ CID 118553325<\/a>
\n\"SCHEMBL17309143\"
\nCompound 9-3
\nSCHEMBL17309143<\/a> \/ CID 118554498<\/a><\/div>\n

Compound 2-2<\/b> was not correctly extracted and looks like OCR has mistakenly recognised the -OBn<\/code><\/strong> as -OBu<\/code><\/strong>. The flurobenzene probably comes from Compound 9-1<\/b> where the label (Boc)2<\/sub>N-<\/code><\/strong> is difficult to recognise. The products of Step 4<\/b> contain valence errors and were probably thrown out as a recognition error.
\nHowever, by reading the ChemDraw files directly it’s possible to extract everything “warts and all”. To process this sketch the key interpretation phases are:<\/p>\n