In an attempt to establish statistically more rigorous and chemically more meaningful quantitative structure-activity relationship (QSAR) equations, a reassessment of a recent study of in vivo anticonvulsant activity for a set of 2,3-benzodiazepin-4(thi)one- and 1,2-phthalazine-related allosteric AMPA antagonists (n = 61) is presented. Contrary to the original, relatively nonspecific descriptor set, which included, for example, a number of topological descriptors, specific structural descriptors that are much easier to interpret from a medicinal chemical point of view are used in this multiple linear regression-based approach. Only statistically significant descriptors have been retained in the final equation, and whereas they give about the same correlation as those of the original paper on the training set (r2 of 0.79 vs. 0.76, n = 49), they perform much better on the test set (predictive r pr 2 of 0.73 vs. 0.05; r2 of 0.78 vs. 0.08, n = 12). Descriptors found to be relevant are clearly related to substitutions at known pharmacophore positions, such as those corresponding to the 2,3-, 7,8- and 4′-positions of the benzodiazepine skeleton. Therefore, by a more careful selection of the descriptor set, both an improved prediction and a more intuitive quantitative interpretation could be achieved for this set of allosteric AMPA antagonists.
- AMPA antagonist
- Linear regression
ASJC Scopus subject areas
- Discrete Mathematics and Combinatorics