Comparison of QSAR Based Thermo and Water Solvation Property Prediction Tools and Experimental Data for Selected Traditional Chemical Warfare Agents and Simulants

摘要

There is a need for rapid prediction of the physico-chemical properties of chemical warfare agents (CWAs) and toxic industrial chemicals (TICs) on environmentally relevant materials, personal protective equipment, and human tissue. It is the objective of this report to survey the reliability from several available in-silico tools for a set of physico-chemical properties that impact prediction of environmental fate of a set of traditional CWAs and simulants. The tools included EPI Suite, ACD Labs, ADF COSMO-RS, ChemAxon's Marvin, and Vega. Of the predictive tools surveyed, EPI Suite and ACD Labs consistently had the highest accuracy for boiling point, vapor pressure. EPI Suite and ACD Labs gave reasonable results for octanol-water partitioning coefficient and water solubility for most compounds evaluated. For available measurements, ACD Labs, COSMO-RS, and Marvin were within two log units of the measured value. Arrangements of atoms outside the model training set accounted for much of the error between prediction and experiment. For molecular properties dependent on descriptors such as dipole moment, the effect of molecular symmetry most likely accounts for significant overestimation. The fragment based models are purely additive and molecular symmetry may cause the effect of a pair of functional groups to cancel rather than add.