Nomenclature translation is the process of converting a human-readable chemical name into a machine-readable notational scheme such as a connection table. It plays a key role in linking the older chemical literature to modern information technologies, such as the Internet.

Buried deep within the Chemistry Development Kit (CDK) is a library for nomenclature translation called ChemNomParse. At the heart of ChemNomParse is a remarkable piece of software called the Java Compiler Compiler (JavaCC), a parser generator and lexical analyzer generator for Java. A FAQ on JavaCC is available here.

This tutorial demonstrates how freely-available, open source tools can be used to parse an IUPAC chemical name and generate its corresponding 2-D structure rendering. A closely-related tutorial on generating 2-D structures from SMILES strings may be helpful as background.

Ingredients

This tutorial uses Arton's Ruby Java Bridge, the installation and use of which has been outlined previously. In addition, you'll need to download Structure-CDK v0.1.2, also previously discussed. Be sure to download v0.1.2, as two upgrades have been released since the package was originally described. This tutorial has been tested on Mandriva Linux 2006.

Create a working directory called nom. From the lib directory of the Structure-CDK distribution, copy cdk-20060714.jar and structure-cdk-0.1.2.jar into your depict working directory.

Code

Create a file called depict.rb and copy the following code into it:

ENV['CLASSPATH'] = './cdk-20060714.jar:./structure-cdk-0.1.2.jar'

require 'rubygems'
require_gem 'rjb'
require 'rjb'

NomParser = Rjb::import 'org.openscience.cdk.iupac.parser.NomParser'
StructureDiagramGenerator = Rjb::import 'org.openscience.cdk.layout.StructureDiagramGenerator'
ImageKit = Rjb::import 'net.sf.structure.cdk.util.ImageKit'

class Depictor

  def initialize
    @sdg = StructureDiagramGenerator.new
  end

  def depict_png(nom, width, height, path_to_png)
    ImageKit::writePNG(nom_to_mol(nom), width, height, path_to_png)
  end

  def depict_svg(nom, width, height, path_to_svg)
    ImageKit::writeSVG(nom_to_mol(nom), width, height, path_to_svg)
  end

private

  def nom_to_mol(nom)
    @sdg.setMolecule(NomParser::generate(nom))
    @sdg.generateCoordinates

    @sdg.getMolecule
  end
end

$ export LD_LIBRARY_PATH=$JAVA_HOME/jre/lib/i386:$LD_LIBRARY_PATH

This tells RJB where to find Java's native libraries. Because of RJB's current design, LD_LIBRARY_PATH needs to be set from the command line, rather than from within a Ruby process.

Using the Depictor class is as simple as creating an instance and invoking depict_png or depict_svg on it:

require 'nom'

depictor = Depictor.new

depictor.depict_png('2-phenylcyclohexan-1-ol', 300, 300, 'output.png')

Executing the above code either through the Ruby interpreter (ruby) or via Interactive Ruby (irb) products a PNG image of the chiral auxiliary shown below:

Other names correctly recognized by ChemNomParse include:

• phenylhexyne
• 2-chloro-3-phenyl-4,4-dimethylhexane
• 3-phenyl-1-aminopropane
• 1,2-difluoro-3-hydroxycyclohexene

Limitations

Many chemical names, ranging from the simple to the complicated, were not be recognized at all by ChemNomParse. Some examples are:

• benzene
• piperidine
• 1-methoxyhexane
• 2-methyl-5-prop-1-en-2-yl-cyclohex-2-en-1-one (carvone)

Some names were incorrectly interpreted due to misassigned locants. For example, 2-chloro-3-hydroxybutanoic acid produced the incorrectly asssigned structure shown below:

ChemNomParse can accurately recognize chemical names representing simple substitutions on basic hydrocarbon scaffolds. More complicated structures, such as heterocycles, bicyclic systems, and systems involving nested substituents do not appear to be handled at all. It is not clear to what extent these limitations reflect a small dictionary of morphemes (the basic nomenclature building blocks) versus deeper design issues.

Despite its limitations, ChemNomParse is an interesting piece of open source software for working with chemical nomenclature. From this simple tutorial, it can be seen that nomenclature translation, when combined with other capabilities such as 2-D rendering, offers many exciting possibilities.

... We report here the development of a computer program for converting chemical names into connection tables, a process we call "nomenclature translation." ... this process provides an alternate method of structure registration by allowing a new substance to be input via a structurally descriptive systematic name instead of only as a connection table taken from a structural diagram.

-G.G.V. Stouw et al. J. Chem. Doc. 1974, 14, 185-193

Systematic nomenclature is one of the oldest forms of line notation. As a result, it can be found widely in papers, patents, spreadsheets, and other documents. Any software that can convert systematic nomenclature, such as IUPAC names, into a computer-based representational system, such as a connection table, has the potential to unlock vast amounts of legacy chemical information by making it structure-searchable.

Stouw and his group at Chemical Abstracts Service (CAS) developed the first working system for name to structure conversion. Their interest in an automated process stemmed from the potential to greatly accelerate the rate at which the chemical literature could be indexed. Instead of a human creating a computer representation by manually parsing a systematic name from a paper, a computer could do it error-free at a fraction of the cost. These factors are still at work today, although the pool of raw chemical information material has increased exponentially since 1974.

Nomenclature translation has been more widely investigated than the related problem of 2-D raster image interpretation, although the driving forces in both cases are the same. There are, of course, several proprietary packages for nomenclature translation. An important disadvantage of all of them is a distinct lack of customizability.

Open source nomenclature translation options have been very limited. One of the first such packages was ChemNomParse by David Robinson, Bhupinder Sandhu, and Stephen Tomkinson at the University of Manchester. ChemNomParse has since been made part of the Chemistry Development Kit (CDK). Although its capabilities are relatively limited, ChemNomParse is very useful for the design it embodies.

More recently, Peter Corbet at Cambridge has developed a package called OPSIN. Egon Willighagen wrote about integrating OPSIN into the desktop software package Bioclipse. OPSIN's source can be found in the project's SVN repository.

The most exciting potential for chemical nomenclature translation is realized when this capability is blended with other chemical informatics technologies. Future articles in this series will show how ChemNomParse and OPSIN can be used with other open source tools to create rich chemical informatics systems.