Prediction of pesticide acute toxicity using two-dimensional chemical descriptors and target species classification

Previous modelling of the median lethal dose (oral rat LD₅₀) has indicated that local class-based models yield better correlations than global models. We evaluated the hypothesis that dividing the dataset by pesticidal mechanisms would improve prediction accuracy. A linear discriminant analysis (LDA) based-approach was utilized to assign indicators such as the pesticide target species, mode of action, or target species - mode of action combination. LDA models were able to predict these indicators with about 87% accuracy. Toxicity is predicted utilizing the QSAR model fit to chemicals with that indicator. Toxicity was also predicted using a global hierarchical clustering (HC) approach which divides data set into clusters based on molecular similarity. At a comparable prediction coverage (~94%), the global HC method yielded slightly higher prediction accuracy (r² = 0.50) than the LDA method (r² ~ 0.47). A single model fit to the entire training set yielded the poorest results (r² = 0.38), indicating that there is an advantage to clustering the dataset to predict acute toxicity. Finally, this study shows that whilst dividing the training set into subsets (i.e. clusters) improves prediction accuracy, it may not matter which method (expert based or purely machine learning) is used to divide the dataset into subsets.     

Rational selection of training and test sets for the development of validated QSAR models

Quantitative Structure–Activity Relationship (QSAR) models are used increasingly to screen chemical databases and/or virtual chemical libraries for potentially bioactive molecules. These developments emphasize the importance of rigorous model validation to ensure that the models have acceptable predictive power. Using k nearest neighbors (kNN) variable selection QSAR method for the analysis of several datasets, we have demonstrated recently that the widely accepted leave-one-out (LOO) cross-validated R² (q²) is an inadequate characteristic to assess the predictive ability of the models [Golbraikh, A., Tropsha, A. Beware of q2! J. Mol. Graphics Mod. 20, 269-276, (2002)]. Herein, we provide additional evidence that there exists no correlation between the values of q ² for the training set and accuracy of prediction (R ²) for the test set and argue that this observation is a general property of any QSAR model developed with LOO cross-validation. We suggest that external validation using rationally selected training and test sets provides a means to establish a reliable QSAR model. We propose several approaches to the division of experimental datasets into training and test sets and apply them in QSAR studies of 48 functionalized amino acid anticonvulsants and a series of 157 epipodophyllotoxin derivatives with antitumor activity. We formulate a set of general criteria for the evaluation of predictive power of QSAR models.     

A Noncongeneric Model for Predicting Toxicity of Organic Molecules to Vibrio Fischeri

Abstract

A general Quantitative Structure-Activity Relationship (QSAR) model on Vibrio fischeri (Microtox^? test) was derived using the autocorrelation method for describing the molecules and a neural network as statistical tool. From a training set of 1068 organic chemicals described by means of four different autocorrelation vectors, it was possible to obtain valuable models but presenting some large outliers. Addition of the time of exposure as variable allowed us to derive a more powerful model from 2795 toxicity results. The predictive power of this 36/26/1 neural network model was tested on an external testing set of 385 toxicity data and compared with the performances of linear models designed for polar narcotic amines and for weak acid respiratory uncouplers.     

Some case studies on application of “rm2” metrics for judging quality of quantitative structure–activity relationship predictions: Emphasis on scaling of response data

Quantitative structure–activity relationship (QSAR) techniques have found wide application in the fields of drug design, property modeling, and toxicity prediction of untested chemicals. A rigorous validation of the developed models plays the key role for their successful application in prediction for new compounds. The r_m² metrics introduced by Roy et al. have been extensively used by different research groups for validation of regression‐based QSAR models. This concept has been further advanced here with introduction of scaling of response data prior to computation of r_m². Further, a web application (accessible from http://aptsoftware.co.in/rmsquare/ and http://203.200.173.43:8080/rmsquare/ ) for calculation of the r_m² metrics has been introduced here. The present study reports that the web application can be easily used for computation of r_m² metrics provided observed and QSAR‐predicted data for a set of compounds are available. Further, scaling of response data is recommended prior to r_m² calculation. © 2013 Wiley Periodicals, Inc.     

The effects of characteristics of substituents on toxicity of the nitroaromatics: HiT QSAR study

The present study applies the Hierarchical Technology for Quantitative Structure–Activity Relationships (HiT QSAR) for (i) evaluation of the influence of the characteristics of 28 nitroaromatic compounds (some of which belong to a widely known class of explosives) as to their toxicity; (ii) prediction of toxicity for new nitroaromatic derivatives; (iii) analysis of the effects of substituents in nitroaromatic compounds on their toxicity in vivo. The 50% lethal dose concentration for rats (LD₅₀) was used to develop the QSAR models based on simplex representation of molecular structure. The preliminary 1D QSAR results show that even the information on the composition of molecules reveals the main tendencies of changes in toxicity. The statistic characteristics for partial least squares 2D QSAR models are quite satisfactory (R ² = 0.96–0.98; Q ² = 0.91–0.93; R ² _test = 0.89–0.92), which allows us to carry out the prediction of activity for 41 novel compounds designed by the application of new combinations of substituents represented in the training set. The comprehensive analysis of toxicity changes as a function of substituent position and nature was carried out. Molecular fragments that promote and interfere with toxicity were defined on the basis of the obtained models. It was shown that the mutual influence of substituents in the benzene ring plays a crucial role regarding toxicity. The influence of different substituents on toxicity can be mediated via different C–H fragments of the aromatic ring. Electronic supplementary material  The online version of this article (doi:) contains supplementary material, which is available to authorized users.

Grid potential analysis,virtual screening studies and ADME/T profiling on N‐arylsulfonylindoles as anti‐HIV‐1 agents

A grid potential analysis employing a novel approach of 3D quantitative structure–activity relationships (QSAR) as AutoGPA module in MOE2009.10 was performed on a dataset of 42 compounds of N‐arylsulfonylindoles as anti‐HIV‐1 agents. The uniqueness of AutoGPA module is that it automatically builds the 3D‐QSAR model on the pharmacophore‐based molecular alignment. The AutoGPA‐based 3D‐QSAR model obtained in the present study gave the cross‐validated Q² value of 0.588, r²_pred value of 0.701, r²_m statistics of 0.732 and Fisher value of 94.264. The results of 3D‐QSAR analysis indicated that hydrophobic groups at R₁ and R₂ positions and electron releasing groups at R₃ position are favourable for good activity. To find similar analogues, virtual screening on ZINC database was carried out using generated AutoGPA‐based 3D‐QSAR model and showed good prediction. In addition to those mentioned earlier, in‐silico ADME absorption, distribution, metabolism and excretion profiling and toxicity risk assessment test was performed, and results showed that majority of compounds from current dataset and newly virtually screened hits generated were within their standard limit. Copyright © 2013 John Wiley & Sons, Ltd.     

Receptor-based 3D QSAR analysis of estrogen receptor ligands – merging the accuracy of receptor-based alignments with the computational efficiency of ligand-based methods

One of the major challenges in computational approaches to drug design is the accurate prediction of binding affinity of biomolecules. In the present study several prediction methods for a published set of estrogen receptor ligands are investigated and compared. The binding modes of 30 ligands were determined using the docking program AutoDock and were compared with available X-ray structures of estrogen receptor-ligand complexes. On the basis of the docking results an interaction energy-based model, which uses the information of the whole ligand-receptor complex, was generated. Several parameters were modified in order to analyze their influence onto the correlation between binding affinities and calculated ligand-receptor interaction energies. The highest correlation coefficient (r ² = 0.617, q ² _LOO = 0.570) was obtained considering protein flexibility during the interaction energy evaluation. The second prediction method uses a combination of receptor-based and 3D quantitative structure-activity relationships (3D QSAR) methods. The ligand alignment obtained from the docking simulations was taken as basis for a comparative field analysis applying the GRID/GOLPE program. Using the interaction field derived with a water probe and applying the smart region definition (SRD) variable selection, a significant and robust model was obtained (r ² = 0.991, q ² _LOO = 0.921). The predictive ability of the established model was further evaluated by using a test set of six additional compounds. The comparison with the generated interaction energy-based model and with a traditional CoMFA model obtained using a ligand-based alignment (r ² = 0.951, q ² _LOO = 0.796) indicates that the combination of receptor-based and 3D QSAR methods is able to improve the quality of the underlying model.