Predicting ecotoxicology of organophosphorous insecticides: Successful parameter selection with the genetic function algorithm

QSARs have been developed for 20 organophosphates, using good experimental data in the form of LC50 toxicity values towards guppy from previously publ ished work. A large number of steric, physical and electronic descriptors h ave been calculated. After the removal of descriptors that were highly cros s-correlated or correlated poorly with activity, the genetic algorithm was employed to select the optimum descriptors for use in multivariate regressi on. Best equation r(2) and r(cv)(2) values were 0.943 and 0.905 respectivel y. Molecular field Analysis was also used in conjunction with the Genetic F unction Approximation (GFA) algorithm to obtain predictive equations, r(2) and r(ev)(2) values were 0.939 and 0.896 respectively. However, when the da ta were randomised, results proved to be as good, (and in one case better) than using the correct experimental values. For equations subsequently deri ved using the Genetic Partial Least Squares (G/PLS) technique (best r(2) an d r(ev)(2) values being 0.943, 0.905 respectively), randomisation gave wors e answers than using the correct experimental values. This work demonstrate s the danger of using the powerful GFA technique to select from a large num ber of parameters, but establishes that the G/PLS method is sufficiently ro bust to provide reliable results.