Multivariate quantitative structure-activity relationship (QSAR) methods which may be applied to pesticide research

A Simple Approach to the Toxicity Prediction of Anilines and Phenols Towards Aquatic Organisms

Chemicals pollution in the environment has attracted attention all over the world, and the toxicity prediction of chemical pollutants has become quite important. In this paper, we introduce a simple approach to predict the toxicity of some chemical components, in which the Tchebichef image moment (TM) method was employed to extract useful chemical information from the images of molecular structures to establish quantitative structure-activity relationship (QSAR) prediction models. The proposed approach was applied to predict the toxicity of anilines and phenols for the aquatic organisms of P. subcapitata and V. fischeri, in which the obtained TMs were defined as the independent variables, while the biological toxicity (pEC₅₀) was regarded to be the dependent variable. Then, the predictive models were established by stepwise regression, respectively. The obtained squared correlation coefficients of leave-one-out cross-validation (Q²) for training sets and the predictive squared correlation coefficients (R_p²) for test sets of the two groups of data were higher than 0.79 and 0.75, respectively, which indicated that the obtained models possessed satisfactory accuracy and reliability. Compared with several reported methods, the proposed approach was more convenient and has a higher predictive capability. Our study provides another perspective in QSAR research.

Modeling antileukemic activity of carboquinones with electrotopological state and chi indices

The antileukemic activity (medium effective dose, MED) of a set of 37 carboquinones was modeled using a combination of the electrotopological state (E-state) and molecular connectivity indices with multiple linear regression. A four-variable model gave good statistics: r2 = 0.90, s = 0.21. Using the leave-one-out method, the cross-validation statistics indicate a model useful for prediction: r2press = 0.85, spress = 0.26. The same variables were used to model the optimum effective dose (OD): r2 = 0.88, s = 0.19. The cross-validation statistics indicate a model useful for prediction: r2press = 0.83, spress = 0.23. The descriptor variables are interpreted in terms of the molecular structure.

Analysis of structure-toxicity relationships for a series of amide herbicides using statistical methods and neural network

Structure-toxicity relationships were studied for a set of 44 herbicides by means of principal component analysis (PCA), multiple regression analysis (MRA), and neural network (NN). The values of log LD50 (lethal dose 50, acute, oral, rat) of the studied compounds were well correlated with the descriptors encoding the chemical structures. Considering the pertinent descriptors, a correlation coefficient of 0.90 (n = 41) was obtained for the NN model with a configuration of 4-3-1 (and 0.92 (n = 41) with a configuration of 4-5-1). To evaluate the contribution of each descriptor on the activity, log LD50 was calculated by removing each descriptor a part. This approach provides the tendency of a descriptor to be favourable (or not) to the activity.

Chiral chromatography and multivariate quantitative structure-property relationships of benzimidazole sulphoxides

Various benzimidazole sulphoxides were chirally resolved employing an amylase-based chiral stationary phase. The structure-property relationships of these compounds were investigated using calculated physicochemical properties, molecular modelling and multivariate statistical techniques. A data set of 254 molecular descriptors was used to represent the series of compounds. Analysis of the data set using principal components analysis and non-linear mapping suggested that the separation factor of each enantiomeric pair was associated with nine molecular properties and, in particular, molar refractivity of the Z substituent and the partial charge of atom 6. The separation factor of a sulphoxide not used in the analysis was well predicted thus suggesting that these models may be used to generalize.

Multivariate quantitative structure-toxicity relationships in a series of dopamine mimetics

The techniques of principal components analysis and non-linear mapping are routinely used by computer chemists at SmithKline Beecham Pharmaceuticals in the process of drug development by relating the structure of a compound to its chemical activity. To our knowledge these techniques had not previously been applied to the association between the structure of a compound and its toxicological properties. Using a series of 12 structurally related compounds (11 were active dopamine mimetics and one was inactive), of which five were known to be teratogenic and seven were non-teratogenic, it was possible to demonstrate that molecular modelling techniques could be applied to differentiate toxicological data. The structure/property relationships of these compounds were investigated using calculated physicochemical properties, molecular modelling and multivariate statistical techniques. A data set of 56 molecular descriptors was used to represent this series of compounds. Analysis of the data set using principal components analysis and non-linear mapping suggested that teratogenicity was associated with four molecular properties. Moreover, the electronic nature of the 4-phenyl group appeared to be an important determinant of the teratogenesis.

Estimating pesticide field half-lives from a backpropagation neural network

The field half-lives of 110 pesticides were modelled using a backpropagation neural network (NN). The molecules were described by means of the frequency of 17 structural fragments. Before training the NN, different scaling transformations were assayed. Best results were obtained with correspondence factor analysis which also allowed a reduction of dimensionality. The training and testing sets of the NN analysis gave 95.5% and 84.6% of good classifications, respectively. Comparison with discriminant factor analysis showed that a backpropagation NN was more appropriate to model the field half-lives of pesticides.

Novel method for the display of multivariate data using neural networks

A neural network has been used to reduce the dimensionality of multivariate data sets to produce two-dimensional (2D) displays of these sets. The data consisted of physicochemical properties for sets of biologically active molecules calculated by computational chemistry methods. Previous work has demonstrated that these data contain sufficient relevant information to classify the compounds according to their biological activity. The plots produced by the neural network are compared with results from two other techniques for linear and nonlinear dimension reduction, and are shown to give comparable and, in one case, superior results. Advantages of this technique are discussed.

Pattern recognition methods in rational drug design

Pattern recognition methods have much to offer the drug designer, particularly as the calculation and collation of data, both biological and physicochemical, becomes easier with the widespread use of computer databases, molecular modeling systems, and property prediction packages. Some of the techniques, however, suffer from difficulties in interpretation and the dangers of chance effects have received little attention. The wider use and understanding of these methods is expected to enhance their utility in drug design. Finally, it should be mentioned here that these methods are becoming applied increasingly in other areas of pharmaceutical research, e.g., the analysis of clinical data, and that new techniques for analysis continue to be developed and applied in this field.