Molecular Property Prediction

Purpose  Suppose you are interested in a compound which has not yet been synthesized or is not yet well studied?  Is it possible to estimate the pKa and aqueous solubility for a potential drug compound?  to estimate the vapor pressure and tendency to partition into fatty tissue for a potential pesticide? 

These questions are not always easy to answer, but the benefits can be substantial.  Reliable estimates can save you the trouble of making (or testing) a compound which is unlikely to meet your needs, or allow you to anticipate how different compounds may behave in a complex system.  Although much of the software in this field is proprietary and expensive, especially for predicting drug interactions with proteins, free software is available to predict a number of simpler properties. 

Recommended    One useful program is EPI-Suite 4.10 from the US Environmental Protection Agency (EPA).  This is the same group of programs recommended for database look-up of physical properties.  EPI-Suite 4.10 (January 2011) can predict boiling point, melting point, vapor pressure, aqueous solubility, octanol-water partitioning, water-air partitioning and environmental degradation rates using the name, CAS number or SMILES structure of a molecule (a video showing how to use SMILES is available for download- see the bottom of this page).  Original references are available for the estimation routines, many of which were developed and/or tested by the Syracuse Research Corporation.  This is a particularly useful program for environmental chemists. EPI-Suite runs only on Windows.



A web-based alternative was developed by the University of Georgia (UGa) in  collaboration with the EPA, and is now maintained by the private company ArcChem.  SPARC (Sparc Performs Automated Reasoning in Chemistry) permits the estimation of the properties above (MP, BP, VP, aqueous solubility, Kow, Henry’s K, etc.) as well as pKa values (including gas phase and non-aqueous phase), density, refractive index, electron affinity and other values.  Note: this site contains JAVA applets which do not run well (or at all) on FireFox 3.5.7 but which perform as expected using Internet Explorer.  Also, some of the prediction routines are still in development and may not always work correctly (the calculations of the properties listed above have worked reliably for me).   

Last updated July 18, 2011

 

Č
č
ď
SMILES_Tutorial.mp4
(3330k)
Steve Cabaniss,
Mar 8, 2010, 12:48 PM
Comments