OSRA: Optical Structure Recognition
1. Description
2. Dependencies
3. Other acknowledgements
4. Compilation
5. Usage
6. License
7. Download
8. Web Interface
9. Author
Description:
OSRA is a utility designed to convert graphical representations of
chemical structures, as they appear in journal articles, patent documents,
textbooks, trade magazines etc., into SMILES (Simplified Molecular
Input Line Entry Specification - see
http://en.wikipedia.org/wiki/SMILES) -
a computer recognizable molecular structure format. OSRA can read a document
in any of the over 90 graphical formats parseable by ImageMagick - including
GIF, JPEG, PNG, TIFF, PDF, PS etc., and generate the SMILES representation of
the molecular structure images encountered within that document.
Note that any software designed for optical recognition is unlikely to
be perfect, and the output produced might, and probably will, contain
errors, so a curation by a human knowledgeable in chemical structures
is highly recommended.
Dependencies:
OSRA needs the following Open Source libraries installed:
- ImageMagick, image manipulation library, version 6.2.8 or later;
if installing from RPM make sure you have the following packages:
ImageMagick
ImageMagick-devel
ImageMagick-c++-devel
ImageMagick-c++
http://www.imagemagick.com/
- POTRACE, vector tracing library, version 1.7 or later,
http://potrace.sourceforge.net/
- GOCR/JOCR, optical character recognition library, version 0.43 or
later - NOTE version 0.46 is NOT RECOMMENDED! It lacks the necessary libPgm2asc library,
http://jocr.sourceforge.net/
- OCRAD, optical character recognition program, version 0.17,
http://www.gnu.org/software/ocrad/ocrad.html
- TCLAP, Templatized C++ Command Line Parser Library, version 1.1.0,
http://tclap.sourceforge.net/
- OpenBabel, open source chemistry toolbox, version 2.2.0 or later;
if installing from RPM make sure you have the following packages:
openbabel
openbabel-devel
http://openbabel.sourceforge.net/wiki/Main_Page
- OR alternatively you can use RDKit as the molecular builder back-end,
version Q32008_1 or later (svn snapshot is recommended at the moment):
http://www.rdkit.org/
Other acknowledgements:
OSRA also makes use of the following software (you do not need to
install it separately, it's included in the distribution):
Compilation:
Unpack downloaded source code for the OCRAD package. Do not compile or install it -
OSRA will automatically patch it and compile the object files it needs.
Compile and/or install all the other necessary dependencies.
Unpack OSRA package.
Edit the included Makefile to make sure you have the correct locations
for potrace, gocr, openbabel (or rdkit), and tclap.
Check that Magick++-config location (it's a script that
comes from ImageMagick installation) is in your PATH. You might have to set
LD_LIBRARY_PATH=$(RDKIT)/bin when using RDKit, or /usr/local/bin when using
OpenBabel.
Set ARCH variable to one of the following: unix - for linux,unix,osx;
win32 - for building on Windows MinGW environment.
Set OPENBABEL_OR_RDKIT to either openbabel or rdkit depending on what toolkit you
have installed.
Running make
should then generate the executable - osra.
Usage:
OSRA can process the following types of images:
- Computer-generated 2D structures, such as found on the PubChem website,
http://pubchem.ncbi.nlm.nih.gov/,
black-and-white and color (use a resolution of 72 dpi),
- Black-and-white PDF and PostScript files, including multi-page ones. Please
note that you need Ghostcript installed for ImageMagick to be able to
parse these kinds of files. OSRA internally renders PS and PDF at a resolution
of 150 dpi, so you do not need to specify the resolution as a commmand-line parameter,
- Scanned images - black-and-white, a resolution of 300 dpi is recommended,
though 150 dpi can also produce fair results. Please make sure the
scanned image is of reasonable quality - an input that's too noisy will
only generate garbage output.
Some common abbreviations, hetero atoms, fused and merged atomic
labels, hash and wedge bonds, and bridge bonds are currently
recognized. Formal charges, isotopes and some element
symbols, i.e. iodine ("I" -- looks too much like a straight line = single
bond), are not.
Command-line options:
./osra --help
will give you a list of available options with short descriptions.
Most common use: ./osra [-r <resolution>] <filename>
Resolution in dpi, default is 300 (unless it's a PS or PDF file as
mentioned above), filename is the name of your image file (or
PS/PDF document).
Other options:
-t, --threshold: Gray level threshold, default is 0.2
for black-and-white images,
-n, --negate: Inverts colors (for white on black images),
-o, --output: Sets a prefix for writing recognized images to files - i.e.
"-o tmp" will create files tmp0.png, tmp1.png... for
each of the structures,
-s, --size: Resize images on output - can be useful for running OSRA
as a backend for a webservice. Example: "-s 300x400".
-g, --guess: Prints out resolution guess when you chose to have automatic
resolution estimate.
-p, --print: Prints out the value of confidence function estimate.
-f, --format: Output format (either smi for SMILES or sdf for SD file format)
License:
This program is free software; the part of the software that was written
at the National Cancer Institute is in the public domain. This does not
preclude, however, that components such as specific libraries used in the
software may be covered by specific licenses, including but not limited
to the GNU General Public License as published by the Free Software Foundation;
either version 2 of the License, or (at your option) any later version;
which may impose specific terms for redistribution or modification.
This program is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307,
USA. See also http://www.gnu.org/.
See the file COPYING for details.
Download:
OSRA is Free and Open Source Software. You are welcome to download
and use it, provided that you understand the terms described above.
Participation in the development is highly encouraged!
osra-1.1.0.tgz -
Added SD file format output, improved wedge bond detection.
Download Windows executable here.
osra-1.0.1.tgz - Minor bug
fixes. OpenBabel-2.2.0 or svn snapshot of RDKit are recommended with this
version.
osra-1.0.0.tgz - Significant
update of the recognition engine. Simplified built instructions.
Please note that the dependencies have changed since the previous
version.
osra-0.9.9.tgz - Build
system upgraded to allow linking to the newer versions of gocr (0.45).
osra-0.9.8.tgz - Added
recognition of old-style aromatic rings with heteroatoms.
osra-0.9.7.tgz - Improved
recognition of color and low-res images.
osra-0.9.6.tgz - Introduced
automatic resolution detection.
osra-0.9.5.tgz - Source code
modified to facilitate compiling with MinGW for Windows platform.
osra-0.9.4.tgz - added
old-style benzene ring recognition
osra-0.9.3.tgz - added
rudimentary formal charge recognition
osra-0.9.2.tgz - improved
handling of hash and wedge bonds
osra-0.9.1.tgz - slightly improved handling of
72dpi color images
osra-0.9.tgz - original
public release
We also welcome your feedback - send us your comments, suggestions,
criticism, or praise to the contact email address below.
Web Interface:
To demonstrate the capabilities (and limitations) of OSRA we have created
the following web interface:
OSRA Web Interface
Try this sample image from the US Patent Office website first:
patent.gif. Use a resolution of 300 dpi.
Author:
Igor Filippov, igorf(AT)helix.nih.gov
2007, SAIC-Frederick, NCI-Frederick, NIH, DHHS, Frederick, MD