A new approach is being developed that optimally combines estimates of different models to produce accurate predictions of chemical properties relevant to environmental analyses and decision makings.  Error correlations among the primary model estimates were obtained from a training chemical set.  These correlations were then used to construct an overall predictor model where the primary estimates were linearly combined.  The resulting linear estimator was applied with the estimates produced by various methods that modeled physicochemical properties based on different quantum�chemical principles.  Preliminary tests showed that the linear estimator approach could significantly improve the estimation accuracy of octanol�water partition coefficient, aqueous solubility, and Henry's Law constant when compared with the primary methods.  This demonstrated that imperfections in existing estimators/methods may be statistically exploited to produce more reliable predictions of important chemical properties.

Combining multiple estimates to improve prediction accuracy of important physicochemical properties relevant to environmental fate and transport analysis

244th National Meeting (2012)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.