Combining unsupervised and supervised artificial neural networks to predictaquatic toxicity

Most quantitative structure-activity relationship (QSAR) models are linear relationships and significant for only a limited domain of compounds. Here we propose a data-driven approach with a flexible combination of unsupervised and supervised neural networks able to predict the toxicity of a large set of different chemicals while still respecting the QSAR postulates. Since QSAR is applicable only to similar compounds, which have similar biological and physicochemical properties, large numbers of compounds are clustered before building local models, and local models are ensembled to obtain the final result. The approach has been used to develop models to predict the fish toxicity of Pimephales promelas and Tetrahymena pyriformis, a protozoan.     

Validation of counter propagation neural network models for predictive toxicology according to the OECD principles: a case study

The OECD has proposed five principles for validation of QSAR models used for regulatory purposes. Here we present a case study investigating how these principles can be applied to models based on Kohonen and counter propagation neural networks. The study is based on a counter propagation network model that has been built using toxicity data in fish fathead minnow for 541 compounds. The study demonstrates that most, if not all, of the OECD criteria may be met when modeling using this neural network approach.     

Validation of counter propagation neural network models for predictive toxicology according to the OECD principles: a case study

The OECD has proposed five principles for validation of QSAR models used for regulatory purposes. Here we present a case study investigating how these principles can be applied to models based on Kohonen and counter propagation neural networks. The study is based on a counter propagation network model that has been built using toxicity data in fish fathead minnow for 541 compounds. The study demonstrates that most, if not all, of the OECD criteria may be met when modeling using this neural network approach.     

QSAR modelling using combined simple competitive learning networks and RBF neural networks

The aim of this study was to propose a QSAR modelling approach based on the combination of simple competitive learning (SCL) networks with radial basis function (RBF) neural networks for predicting the biological activity of chemical compounds. The proposed QSAR method consisted of two phases. In the first phase, an SCL network was applied to determine the centres of an RBF neural network. In the second phase, the RBF neural network was used to predict the biological activity of various phenols and Rho kinase (ROCK) inhibitors. The predictive ability of the proposed QSAR models was evaluated and compared with other QSAR models using external validation. The results of this study showed that the proposed QSAR modelling approach leads to better performances than other models in predicting the biological activity of chemical compounds. This indicated the efficiency of simple competitive learning networks in determining the centres of RBF neural networks.     

Analysis of the internal representations developed by neural networks for structures applied to quantitative structure--activity relationship studies of benzodiazepines

An application of recursive cascade correlation (CC) neural networks to quantitative structure-activity relationship (QSAR) studies is presented, with emphasis on the study of the internal representations developed by the neural networks. Recursive CC is a neural network model recently proposed for the processing of structured data. It allows the direct handling of chemical compounds as labeled ordered directed graphs, and constitutes a novel approach to QSAR. The adopted representation of molecular structure captures, in a quite general and flexible way, significant topological aspects and chemical functionalities for each specific class of molecules showing a particular chemical reactivity or biological activity. A class of 1,4-benzodiazepin-2-ones is analyzed by the proposed approach. It compares favorably versus the traditional QSAR treatment based on equations. To show the ability of the model in capturing most of the structural features that account for the biological activity, the internal representations developed by the networks are analyzed by principal component analysis. This analysis shows that the networks are able to discover relevant structural features just on the basis of the association between the molecular morphology and the target property (affinity).     

QSAR modeling and design of cationic antimicrobial peptides based on structural properties of amino acids

Drug resistance to existing antibiotics poses alarming threats to global public health, which inspires heightened interests in searching for new antibiotics, including antimicrobial peptides (AMPs). Accurate prediction of antibacterial activities of AMPs may expedite novel AMP design and reduce the costs and efforts involved in laboratory screening. In the present study, a novel quantitative prediction method of AMP was established by quantitative structure-activity relationship (QSAR) modeling based on the physicochemical properties of amino acids. The indices of these physicochemical properties were used to define AMP. The structural variables were optimized by stepwise regression (STR). Three series of AMPs from the QSAR model were constructed by multiple linear regressions (MLR). These QSAR models showed good performance in reliability and predictability. The normalized regression coefficients of the QSAR model and the contribution of amino acids at each position of AMP may determine the suitableness of a particular residue at any given position. QSAR models constructed by STR-MLR should prove to be useful tools in peptide design with respect to the calculation, explanation, good and reliable performance, and definition of physiochemical properties.     

A Quantitative Structure–Activity Relationship Study of Calpeptin (Calpain Inhibitor) as an Anticancer Agent

Calpeptin analogs show anticancer properties with inhibition of calpain. In this work, we applied a quantitative structure–activity relationship (QSAR) model on 34 calpeptin derivatives to select the most appropriate compound. QSAR was employed to generate the models and predict the more significant compounds through a series of calpeptin derivatives. The HyperChem, Gaussian 09, and Dragon software programs were used for geometry optimization of the molecules. The 2D and 3D molecular structures were drawn by ChemDraw (Ultra 16.0) and Chem3D (Pro16.0) software. The Unscrambler program was used for the analysis of data. Multiple linear regression (MLR‐MLR), partial least‐squares (MLR‐PLS1), principal component regression (MLR‐PCR), a genetic algorithm‐artificial neural networks (GA‐ANN), and a novel similarity analysis‐artificial neural network (SA‐ANN) method were used to create QSAR models. Among the three MLR models, MLR‐MLR provided better statistical parameters. The R² and RMSE of the prediction were estimated as 0.8248 and 0.26, respectively. Nevertheless, the constructed model using GA‐ANN revealed the best statistical parameters among the studied methods (R² test = 0.9643, RMSE test = 0.0155, R² train = 0.9644, RMSE train = 0.0139). The GA‐ANN model is found to be the most favorable method among the statistical methods and can be employed for designing new calpeptin analogs as potent calpain inhibitors in cancer treatment.     

Pesticides as estrogen disruptors: QSAR for selective ERα and ERβ binding of pesticides

Evidence suggests that environmental exposure to estrogen-like compounds can cause adverse effects in humans and wildlife. The Endocrine Disruptor Screening and Testing Advisory Committee (EDSTAC) has advised screening of 87,000 compounds in the interest of human safety. This may best be accomplished by pre-screening using quantitative structure-activity relationship (QSAR) modelling. The present study aimed to develop in silico QSARs based on natural, semi-synthetic, synthetic, and phytoestrogens, to predict the potential estrogenic toxicity of pesticides. A diverse set of 170 compounds including steroidal-, synthetic- and phytoestrogens, as well as pesticides was used to construct the QSAR models using artificial neural networks (ANNs). Mean correlation coefficients between experimentally measured and predicted binding affinities were all greater than 0.7 and models had few false negative results, an important consideration for screening tools. This study demonstrated the utility of ANNs as QSAR models for pre-screening of potential endocrine disruptors.     

Tuning neural and fuzzy-neural networks for toxicity modeling

The need for general reliable models for predicting toxicity has led to the use of artificial intelligence. We applied neural and fuzzy-neural networks with the QSAR approach. We underline how the networks have to be tuned on the data sets generally involved in modeling toxicity. This study was conducted on 562 organic compounds in order to establish models for predictive the acute toxicity in fish.     

Use of automatic relevance determination in QSAR studies using Bayesian neural networks

We describe the use of Bayesian regularized artificial neural networks (BRANNs) coupled with automatic relevance determination (ARD) in the development of quantitative structure-activity relationship (QSAR) models. These BRANN-ARD networks have the potential to solve a number of problems which arise in QSAR modeling such as the following: choice of model; robustness of model; choice of validation set; size of validation effort; and optimization of network architecture. The ARD method ensures that irrelevant or highly correlated indices used in the modeling are neglected as well as showing which are the most important variables in modeling the activity data. The application of the methods to QSAR of compounds active at the benzodiazepine and muscarinic receptors as well as some toxicological data of the effect of substituted benzenes on Tetetrahymena pyriformis is illustrated.     

Qsar Models for Predicting the Acute Toxicity of Selected Organic Chemicals with Diverse Structures to Aquatic Non-Vertebrates and Humans

Abstract

The linear and non-linear relationships of acute toxicity (as determined on five aquatic non-vertebrates and humans) to molecular structure have been investigated on 38 structurally-diverse chemicals. The compounds selected are the organic chemicals from the 50 priority chemicals prescribed by the Multicentre Evaluation of In Vitro Cytotoxicity (MEIC) programme. The models used for the evaluations are the best combination of physico-chemical properties that could be obtained so far for each organism, using the partial least squares projection to latent structures (PLS) regression method and backpropagated neural networks (BPN). Non-linear models, whether derived from PLS regression or backpropagated neural networks, appear to be better than linear models for describing the relationship between acute toxicity and molecular structure. BPN models, in turn, outperform non-linear models obtained from PLS regression. The predictive power of BPN models for the crustacean test species are better than the model for humans (based on human lethal concentration). The physico-chemical properties found to be important to predict both human acute toxicity and the toxicity to aquatic non-vertebrates are the n?octanol water partition coefficient (Pow) and heat of formation (HF). Aside from the two former properties, the contribution of parameters that reflect size and electronic properties of the molecule to the model is also high, but the type of physico-chemical properties differs from one model to another. In all of the best BPN models, some of the principal component analysis (PCA) scores of the ¹³C-NMR spectrum, with electron withdrawing/accepting capacity (LUMO, HOMO and IP) are molecular size/volume (VDW or MS1) parameters are relevant. The chemical deviating from the QSAR models include non-pesticides as well as some of the pesticides tested. The latter type of chemical fits in a number of the QSAR models. Outliers for one species may be different from those of other test organisms.     

Parsimonious Optimization of Multitask Neural Network Hyperparameters

Neural networks are rapidly gaining popularity in chemical modeling and Quantitative Structure–Activity Relationship (QSAR) thanks to their ability to handle multitask problems. However, outcomes of neural networks depend on the tuning of several hyperparameters, whose small variations can often strongly affect their performance. Hence, optimization is a fundamental step in training neural networks although, in many cases, it can be very expensive from a computational point of view. In this study, we compared four of the most widely used approaches for tuning hyperparameters, namely, grid search, random search, tree-structured Parzen estimator, and genetic algorithms on three multitask QSAR datasets. We mainly focused on parsimonious optimization and thus not only on the performance of neural networks, but also the computational time that was taken into account. Furthermore, since the optimization approaches do not directly provide information about the influence of hyperparameters, we applied experimental design strategies to determine their effects on the neural network performance. We found that genetic algorithms, tree-structured Parzen estimator, and random search require on average 0.08% of the hours required by grid search; in addition, tree-structured Parzen estimator and genetic algorithms provide better results than random search.     

Variable selection for QSAR by artificial ant colony systems

Derivation of quantitative structure-activity relationships (QSAR) usually involves computational models that relate a set of input variables describing the structural properties of the molecules for which the activity has been measured to the output variable representing activity. Many of the input variables may be correlated, and it is therefore often desirable to select an optimal subset of the input variables that results in the most predictive model. In this paper we describe an optimization technique for variable selection based on artificial ant colony systems. The algorithm is inspired by the behavior of real ants, which are able to find the shortest path between a food source and their nest using deposits of pheromone as a communication agent. The underlying basic self-organizing principle is exploited for the construction of parsimonious QSAR models based on neural networks for several classical QSAR data sets.     

Chemometrical analysis of computed QSAR parameters and their use in biological activity prediction

This study gives a quantitative structure-activity relationship (QSAR) correlation of the 72 N-benzylsalicylamide derivatives properties with their antimycobacterial activity. The antimycobacterial activity was measured as the minimal inhibition concentration (MIC) determined for four strains of mycobacterium (M. avium, M. kansasii, M. kansasii clin.-clinically isolated form, and M. tuberculosis) after 14 days and after 21 days of cultivation. The objective was to identify the factors most closely defining biological activity of N-benzylsalicylamides, in order to enable QSAR prediction of new derivatives with high antimycobacterial activity. Optimal properties for the QSAR analysis were selected from several physicochemical properties, including lipophilicity parameter log P, molecular mass M, molar refraction MR, NMR chemical shifts, polarizability, etc. Many of the considered properties are different from those typically used in traditional QSAR. Selection of the most important properties was performed by one-way Analysis of Variance (ANOVA) and correlation analysis using the significance coefficients and the correlation coefficients, respectively. The chosen variables were further used in artificial neural networks (ANN) for predicting biological activity in the form of-log(MIC). Presented at the 1st International Conference “Applied Natural Sciences” on the occasion of the 10th anniversary of the University of Ss. Cyril and Methodius, Trnava, 7–9 November 2007.