The Promise and the Reality of Turning Chemical Literature into
Information
Tom DomanChicago ACS – Wiggins Symposium
3/25/2007
Abstract• The fields of computational chemistry and
especially chemical informatics have advanced greatly in recent years, and vastly more sophisticated analyses of chemical results are possible these days. However, due to technical, business, and political realities, there are still significant barriers to conducting these analyses in a timely fashion. Using a recent case study, I will highlight the challenges in turning traditional 2D chemical graphs found in many chemistry-related publications into 3D molecular models.
Opportunities Provided by Gary Wiggins:
• Indiana University School of Informatics, IUPUI: Informatics 572 – Computational Chemistry and Molecular Modeling
• External Advisor – NIH funded Exploratory Center for Cheminformatics Research at IU
• Helpful conduit for communication & collaboration between Lilly & IU
Acc. Chem. Res. 2006, 39, 539-549.
Natural Products Discussed in Danishefsky Paper (p. 540)
Can Cheminformatics / Molecular Modeling Shed Some Light on 3D Similarity of These Compounds?
1. Retrieve structures (preferably from public / free sources)
2. Convert to 3D (if necessary)3. Generate conformers4. Perform 3D similarity / pharmacophore
analysis
The Desired Transformation
Easy, Right?• Google• Wikipedia• PubChem• emolecules• DNP on the Web• SciFinder• Corina• WebLab ViewerLite• Catalyst
Google“tricycloillicinone”
• 89 hits• None seem to lead to structure in
electronic form
Wikipedia
• No entries found for any of the 7:– Tricycloillicinone– Merrilactone A– Jiadifenin– Scabronine G– Garsubellin A– 11-O-Debenzoyltashironin– NGA0187
emolecules.com
• Alas, no entries for any of the 7 names (or carefully chosen snippets of them)
DNP on the Web
• Entries found for:– Tricycloillicinone– Jiadifenin
• Registration for Free Trial
SciFinder
• Structures found for all compounds except NGA0187
Further Processing• SMILES generated, but Corina failed on 4• So, from SciFinder, viewed in WebLab ViewerLite &
converted to MOL files• Imported as is to Accelrys “Catalyst”• No conformational model could be built for Jiadifenin, but
others okay• Many HipHop (common features) models built, shown on
next slide is from:– Garsubellin A, Merrilactone A, Scabronine G– Permitted:
• 0-5 HBA• 0-5 Hydrophobic• 0-5 Hydrophobic Aliphatic
Potentially Interesting Overlay of Garsubellin A (yellow) & Scabronine G (white)
Conclusions
• Chemical structure still sometimes hard to obtain from open sources.
• Natural product stereochemistry may stymie computational tools.
• Manuscripts typically don’t embody live structures (but wait, it’s 2007!); here must return to carefully curated structure.
• A simple 2D3D problem ends up being pretty complicated!