You are here:
COREPA-M: NEW MULTI-DIMENSIONAL FUNCTIONALITY OF THE COREPA METHOD
Citation:
Mekenyan, O. G., N. Nikolova, P. K. Schmieder, S P. Bradbury, AND G D. Veith. COREPA-M: NEW MULTI-DIMENSIONAL FUNCTIONALITY OF THE COREPA METHOD. Presented at 10th International Workshop on Quantitative Structure-Activity Relationships (QSARs) in Environmental Sciences, Ottawa, Ontario, Canada, May 25-29, 2002.
Description:
The COmmon REactivity PAttern (COREPA) method is a recently developed pattern recognition technique accounting for conformational flexibility of chemicals in 3-D quantitative structure-activity relationships (QSARs). The method is based on the assumption that non-congeneric chemicals, which elicit similar biological behavior through the same mechanism of action, should possess a commonality in their steric and/or electronic structure, thus yielding a common reactivity pattern. All energetically reasonable conformers of a chemical are considered in deriving this pattern. They are plotted across a molecular descriptor axis, thus forming a discrete distribution for the chemical relative to the selected descriptor. Further, each parameter point value is considered to be the midpoint of a continuous probability distribution: accomplished using various probabilistic density functions (called "kernels"). The conformer distribution of a chemical, or classes of biologically similar chemicals, is obtained as a normalized sum (over the number of conformers) of these probability distributions, ensuring the area of the resulting distribution is unity and hence is considered as a probability density function. Well defined or distinct reactivity patterns are observed when the conformer distributions for the individual chemicals from a class are harmonized (in phase). A variation of the Kernel density formula can be used to account for conformer multiplicity of chemicals, where conformers are weighted using Boltzman statistics. In the original version of COREPA, the probabilistic distributions are estimated across single stereoelectronic parameters. In the present formulation, a multi-dimensional reactivity pattern (COREPA-M) is determined which will enhance discriminating classes of biologically dis-similar chemicals, and builds a decision tree for their identification, maximizing prediction sensitivity (or other goodness criteria).Applicability of COREPA-M is illustrated for predicting chemical modes of toxic action.