A recent series of D-F articles have discussed the use of NestedVM to compile cheminformatics programs written in C/C++ to pure java binaries that can be run on any system with a JVM. More specifically, an attempt to compile OpenBabel's babel program to bytecode was only partially successful. With the help of Geoff Hutchison, the problem was resolved. This article introduces JBabel, a platform-independent, pure Java implementation of OpenBabel's babel program.

A Little About JBabel

JBabel was compiled from the Open Babel 2.1.1 source release and can be downloaded from SourceForge. The same jarfile was successfully tested on Linux, Windows and Mac OS X. You can verify JBabel works on your platform with the following command:

$ java -jar jbabel-20071209.jar -Hsmi
smi  SMILES format
A linear text format which can describe the connectivity
and chirality of a molecule
Write Options e.g. -xt
  n no molecule name
  t molecule name only
  r radicals lower case eg ethyl is Cc

This version of JBabel was compiled with support for three formats:

• SMILES (smi). Non-canonical SMILES.

• MDL (mol). Molfiles and SD Files.

• Canonical SMILES (can). Canonical SMILES implementation donated by eMolecules.

I'll discuss exactly how support for these formats was added in a subsequent post. More formats will be added in the future. For now, let's just try JBabel out.

Testing JBabel

One way to use JBabel is interactively from the command line - just leave out an input or output file parameter. For example, if you wanted to get the eMolecules canonical SMILES for sertraline, you might do something like this (be sure to use two returns to begin processing):

$ java -jar jbabel-20071209.jar -ismi -ocan
CN[C@H]1CC[C@H](C2=CC=CC=C12)C3=CC(=C(C=C3)Cl)Cl

CN[C@H]1CC[C@H](c2ccc(Cl)c(Cl)c2)c2ccccc12
1 molecule converted
34 audit log messages

This canonical SMILES can be converted into a molfile with the following:

$ java -jar jbabel-20071209.jar -ismi -omol
CN[C@H]1CC[C@H](c2ccc(Cl)c(Cl)c2)c2ccccc12


 OpenBabel12090723182D

 22 24  0  0  0  0  0  0  0  0999 V2000
    0.0000    0.0000    0.0000 C   0  0  0  0  0

...

To convert using input and output files, we could use a medium-sized dataset such as the PubChem benzodiazepine dataset prepared for Rubidium:

$ java -jar jbabel-20071209.jar -imol pubchem_benzodiazepine_20071110.sdf -ocan pubchem_benzodiazepine_20071110.smi
==============================
*** Open Babel Warning  in ReadMolecule
  WARNING: Problems reading a MDL file
Cannot read title line

2117 molecules converted

This test, which parses 2117 records, required four minutes forty-five seconds on my system. For comparison, the natively compiled binary did the same thing in about thirteen seconds. Clearly, the JBabel performance hit is substantial.

Uses

Although it's very unlikely that JBabel will ever be useful in performance-critical situations, its portability makes it attractive for other uses. Examples include:

• application development in heterogeneous computing environments;

• use on systems in which native compilation may be difficult, such as those with unusual configurations or operating systems;

• cases in which native binaries work poorly or not at all, such as in applets and Java applications;

• situations in which performance is a minor consideration, such as in end-user applications that process only a few molecules at a time, or during application prototyping

Conclusions

This article has described JBabel, the first portable binary version of OpenBabel's babel molecular file format interconversion program. The next article in this series will describe in detail the steps that were used to compile it.

A very common operation in cheminformatics is the interconversion of molfiles and SMILES strings. Usually, converting from SMILES gives a molfile in which all atoms have coordinates of (0,0,0). Sometimes you just need more than that. The following Ruby CDK code will accept an arbitrary SMILES string and return a molfile with fully-assigned 2D atom coordinates:

require 'rubygems'
require 'rcdk'
require 'rcdk/util'
include RCDK::Util

XY.coordinate_molfile Lang.smiles_to_molfile('c1ccccc1')

Looking at it this way, those four lines of require/include statements seem pretty darn verbose.

Chemists are accustomed to embedding live molecular objects in their documents with Microsoft Word/ChemDraw. These objects can then be reprocessed and embedded into other documents, such as PowerPoint presentations, saving enormous amounts of time. What if the same feature were available with Web documents?

A recent D-F article proposed a method to encode molecular structure data within commonly-used Web image formats such as PNG. That article contained an embedded image of GlaxoSmithKline's diabetes treatment rosiglitazone (Avandia) encoded by a rendering toolkit built with Firefly. I claimed that this image contained the complete connection table and atom coordinates as embedded metadata. In this article, I'll show a simple method to read this metadata.

Metadata is a standard part of the PNG specification; to read it requires nothing more than software capable of recognizing it. I recently found a Web-based, cross-platform method for doing so. The Image Metadata Viewer by FileFormat.info accepts an uploaded image file and returns that image's metadata. Let's try it with the image of rosiglitazone.

After saving the image to my hard drive, uploading it to FileFormat.info and pressing start, I can see that the image contains metadata:

The metadata can be viewed either as XML or as plain text. Choosing plain text (second option) gives me the complete molfile, stored as a key/value hash (molfile=[molfile]).

Clearly, reading metadata is not a problem given the right software. But this leaves the question of how metadata is encoded in the first place - especially in a programming language such as Java. Like everything else, it's not difficult when you know how. Stay tuned for the answer.

The graphical language of 2D structures has served chemistry well for the last 100 years. Ironically, this language which is so useful for human communication is extraordinarily difficult for machines to understand. Heroic efforts at digital raster image recognition such as OSRA and those recently summarized by Egon Willighagen, in addition to a handful of others, have tried to tackle this problem with varying degrees of success.

The problem remains unsolved, and continues to be one of the most difficult technical challenges in cheminformatics. But the pace at which non-machine readable images are generated has accelerated dramatically in the last two years with the emergence of numerous free chemical databases.

What if 2D structure images simply contained all of the information needed for machine processing in the first place?

This idea isn't as far-fetched as it may sound initially. As discussed in a recent D-F article, both GChemPaint and ACD ChemSketch have been claimed to be capable of encoding machine-readable structure information.

Previous D-F articles have described "Firefly", the codename for a new lightweight 2D structure editor designed specifically for the Web. With major work on the editor's user interface complete, more recent efforts have focused on implementing a 2D rendering toolkit, and with it a mechanism to encode structural information within 2D molecular images.

As a demonstration of what is now possible, consider the structure of GlaxoSmithKline's diabetes treatment rosiglitazone (Avandia), depicted as a PNG image at the beginning of this article. At first glance, the image appears to be just like any other image of a 2D molecular structure. But it is not, for embedded within it are the connection table and 2D atom coordinates of rosiglitazone encoded as an industry-standard molfile.

Given the right software, a computer can interpret the structural information encoded in the rosiglitazone image and precisely re-create the original molecular representation. A graphical diagnostic tool bundled with Firefly was equipped with code for precisely this purpose.

This tool can work with molfile-encoded PNG images just as easily as it can with molfiles; they can be opened and the resulting molecule can be further edited, saved in another format, or re-written as a embedded-molfile PNG image.

The first step is to select the PNG image from a local hard drive:

Opening this image produces a fully-editable version of the original molecule:

Obviously, nothing limits this technique to molfiles. InChI, SMILES, CML, or any other molecular encoding scheme would work just as well.

Using molecular-encoded PNG images as a Web-ready replacement for the Word/Chemdraw OLE technology may be one application of this approach. With a large corpus of these images, chemical Web spidering and data mining would be possible on a scale unimaginable today. As always, these possibilities reinforce the desperate need for high quality tools that chemists actually want to use, and which simultaneously yield machine-readable output.

If applications are hard to design and toolkits are harder, then frameworks are hardest of all. A framework designer gambles that one architecture will work for all applications in the domain. Any substantive change to the framework's design would reduce its benefits considerably, since the framework's main contribution to an application is the architecture it defines. Therefore it's imperative to design the framework to be as flexible and extensible as possible.

-Erich Gamma, Richard Helm, Ralph Johnson, and John Vlissides- Design Patterns

One of the most important considerations when building an application is the choice of framework. As the quote from the Gang of Four implies, there's much more to frameworks than just a collection of re-usable code. At their best, frameworks provide a foundation for thinking about a problem domain and a language for communicating with other developers about it. In this article, I'll introduce Octet, an object-oriented framework for molecular representation.

The Molecular Representation Problem

Isn't molecular representation a solved problem? After all, don't SMILES, Molfile, InChI, and CML adequately represent any molecule the average software developer is likely to see?

As previously discussed, molecular representation technologies have stagnated while the molecules chemists themselves now routinely make and use have continued to become more and more "exotic." Developers are now faced with the thorny problem that a variety of common structural motifs in chemistry can't be adequately represented with industry-standard cheminformatics tools.

This point is so important, I'll repeat it: cheminformatics has fallen behind chemistry in the kinds of molecules it can work with. Quick fixes only allow the problem to fester; what's needed is a comprehensive solution. This is Octet's problem domain.

Every framework is bounded by a specific problem domain. Although the size of the domain can vary, a framework provides a comprehensive solution within it. For complex and poorly standardized domains (such as molecular representation), a good framework can greatly accelerate application development.

A good frameworks stays within its problem domain. One of the most important reasons is to prevent featuritis, the root of much software evil. Keeping a framework focused on its core mission makes it much more likely that it can remain documented, tested, extensible, and efficient.

By intention, a variety of features fall outside Octet's problem domain and so will never be directly supported. For example, rendering 2-D structure diagrams is a common problem in cheminformatics that has nothing to do with solving the molecular representation problem. Similarly, reading and writing SMILES strings and Molfiles are supported by many toolkits, but not by Octet directly. After all, it's the inherent limitations of these languages that Octet is trying to overcome.

Higher-level functionality such as legacy language support and 2-D rendering, although not part of Octet itself, can be developed with Octet as a foundation. For example, two Octet add-on frameworks specifically address these problems. They are called Rosetta and Structure, respectively.

About This Series

This article is the first in a series discussing Octet. Future articles will describe in detail Octet's design, implementation, and use. Although Octet has come a long way, it's far from finished. My motivation for writing these articles is to hear what you have to say about Octet, so please feel free to contact me.

Although Octet is written in Java, code examples discussed here will be written in Ruby. I've taken the same approach in discussing the Chemistry Development Kit (CDK) and Structure-CDK. Ruby's brevity and comfortable syntax make it ideal for both writing and discussing code.

Ruby Java Bridge (RJB) is the magic technology that makes this possible. Previous articles have discussed the installation and use of RJB on Windows and Linux.

A Simple Test

Assuming you've installed Ruby, RubyGems and Ruby Java Bridge, you can perform a simple demonstration of Octet in Ruby:

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

BasicMoleculeBuilder = Rjb::import('net.sf.octet.builder.BasicMoleculeBuilder')
RepresentationKit = Rjb::import('net.sf.octet.util.RepresentationKit')
MoleculeKit = Rjb::import('net.sf.octet.util.MoleculeKit')
System = Rjb::import('java.lang.System')

builder = BasicMoleculeBuilder.new

RepresentationKit.buildHexane(builder)

molecule = builder.releaseMolecule

MoleculeKit.printMolecule(molecule, System.out)

$ export CLASSPATH=./octet-0.8.2.jar
$ ruby test.rb
**Molecule Properties**

Atom Count: 6, Bonding System Count: 5

Atoms:
atom: C[0] (2nu 0e, 0or, 0.0fc, 1bs, 1n, 4.0val, 3ih )
atom: C[1] (2nu 0e, 0or, 0.0fc, 2bs, 2n, 4.0val, 2ih )
atom: C[2] (2nu 0e, 0or, 0.0fc, 2bs, 2n, 4.0val, 2ih )
atom: C[3] (2nu 0e, 0or, 0.0fc, 2bs, 2n, 4.0val, 2ih )
atom: C[4] (2nu 0e, 0or, 0.0fc, 2bs, 2n, 4.0val, 2ih )
atom: C[5] (2nu 0e, 0or, 0.0fc, 1bs, 1n, 4.0val, 3ih )

No non-natural isotopic distributions specified.

No Orbitals specified.

Bonding Systems:
bonding system:  ( 2be, 0abe, 2a, 1ap ) [ (0, 1) ]
bonding system:  ( 2be, 0abe, 2a, 1ap ) [ (1, 2) ]
bonding system:  ( 2be, 0abe, 2a, 1ap ) [ (2, 3) ]
bonding system:  ( 2be, 0abe, 2a, 1ap ) [ (3, 4) ]
bonding system:  ( 2be, 0abe, 2a, 1ap ) [ (4, 5) ]

Atom Pairs:
atom pair: (0, 1) (1.0 bo)
atom pair: (1, 2) (1.0 bo)
atom pair: (2, 3) (1.0 bo)
atom pair: (3, 4) (1.0 bo)
atom pair: (4, 5) (1.0 bo)

No Atomic Configurations specified.
No Conformation specified.

As you can see, Octet shares the same concepts and vocabulary as FlexMol. We'll drill down into the meaning of the output in later articles. The important thing to remember is that we can print out a report like the one above for any Molecule, no matter how complex.

Conclusions

Octet is an object-oriented framework designed to solve the molecular representation problem and serve as a solid foundation for a variety of cheminformatics applications. Of course, there's much more to Octet than the simple example shown here. Future articles will describe in greater detail the design and use of Octet through illustrative examples.