Depth-First: Structure Diagram Generation

Structure Diagram Generation

Rich Apodaca — Wed, 11 Apr 2007 10:20:00 +0000

Given a molecule with no 2D coordinates, how would you render a human-readable view? This problem arises in many situations, but most commonly in the context of interpreting line notations such as IUPAC nomenclature, SMILES, or InChI. Whatever the solution you come up with, you'll come face-to-face with the structure diagram generation (SDG) problem.

Generating 2D molecular coordinates is a fundamental (and remarkably difficult) problem in cheminformatics. Discussions in the primary literature date back to at least the 1970s with Chemical Abstract Service's pioneering large-scale efforts. A recent article from Chemical Computing Group (CCG) described the design and implementation of an advanced SDG system. To my knowledge, the only open source implementation of an SDG system is found in the Chemistry Development Kit, and by extension Ruby CDK.

The SDG problem plays an important role in the aesthetics of chemical structure diagrams, as mentioned by two readers. To render a molecule aesthetically, 2D coordinates must minimize confusing atom overlaps, unconventional orientations, and unusual bond angles.

The role of SDG in cheminformatics can only continue to increase in importance, especially as more and more structures are automatically generated through mining the primary literature, the Internet, old PDFs, and other sources. With all of these new computer-generated structures will come the need to make them readily understandable to a chemist through SDG.

"Structure Diagram Generation" by Rich Apodaca

Sun, 22 Apr 2007 01:03:24 +0000

Thanks, Geoff - I had missed that. Looks like the MCDL editor does do structure diagram generation. As a bonus the MCDL source is public domain.

"Structure Diagram Generation" by Geoff

Mon, 16 Apr 2007 15:35:11 +0000

Wasn't there another open source 2D layout. I remember it was at the fall 2006 ACS meeting in San Francisco in the CINF general session.

Wait, I found it: http://sourceforge.net/projects/mcdl

It's public domain. Their line notation is a little strange, but it might be interesting to look at their source.

"Structure Diagram Generation" by Rich Apodaca

Sat, 14 Apr 2007 10:56:39 +0000

Ah yes, MDRAW. I also remember the pains I went to many years ago to track down that source code on microfiche. Although the source code is there to read, if I'm not mistaken it's copyrighted by J. Chem. Res. The code itself has not, to my knowledge, been released under an open source license. So it's free beer, but not free speech.

"Structure Diagram Generation" by Christoph Steinbeck

Thu, 12 Apr 2007 05:20:24 +0000

In fact, there is one more open source implementation:

K Bley, J Brandt, A Dengler, R Frank, and I Ugi (1991) Constitutional Formulae generated from Connectivity Information: the Program MDRAW. Journal of Chemical Research (M):2601-2689.

This is a funny one. The whole source code what published in print form in the microfiche edition of the Journal of Chemical Research. The print edition had only the article. In 1995, I got very bad print copy of the micro fiche by interlibrary loan and my daddy and I sat down and typed in the MDRAW C code. It was horrible. I never got it to run and later managed to contacted Alf Dengler, who sent me the MDRAW C sources. Those formed the basis of what I used in LUCY to display results. MDRAW was designed to produce diagrams with all carbons symbols drawn and therefore produced linear carbon chains. In 2002 I wrote the CDK structure diagram generation completely from scratch - no concepts from MDRAW where used.