Structure−Activity Relationship Studies of Carcinogenic Activity of Polycyclic Aromatic Hydrocarbons Using Calculated Molecular Descriptors with Principal Component Analysis and Neural Network Methods

Vendrame, Rosana; Braga, Rafael Silveira; Takahata, Yuji; Galvão, Douglas S.

doi:10.1021/ci990326v

Cited by 62 publications

(58 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Multilayer feed-forward neural networks trained using back-propagation learning algorithms have become increasingly popular [28][29][30]. Principal component-artificial neural network (PC-ANN) combines the PCA and ANN methods and models the nonlinear relationships between the PCs and a dependent variable [31,32].…”

Section: Introductionmentioning

confidence: 99%

Comparative QSRR Modeling of Nitrobenzene Derivatives Based on Original Molecular Descriptors and Multivariate Image Analysis Descriptors

Garkani-Nejad

Ahmadvand

2011

Chromatographia

View full text Add to dashboard Cite

A comparative quantitative structure-retention relationship (QSRR) study has been carried out to predict the retention time of nitrobenzene derivatives using original molecular descriptors and multivariate image analysis (MIA) descriptors. First, original molecular descriptors were generated from molecular structures and applied to construct QSRR models using multiple linear regression (MLR), partial least squares (PLS), and artificial neural network (ANN) modeling methods. Then, multivariate image analysis (MIA) descriptors were generated from pixels of images and analyzed using correlation rankingprincipal component regression (CR-PCR) and correlation ranking-principal component-artificial neural network (CR-PC-ANN) methods. In this paper, the CR-PC-ANN method presented better results than the other methods for predicting the retention time of the studied compounds. Coefficients of determination (R 2 ) using the CR-PC-ANN method for the training, test, and validation sets were 0.989, 0.999, and 0.999, respectively.

show abstract

Section: Introductionmentioning

confidence: 99%

Comparative QSRR Modeling of Nitrobenzene Derivatives Based on Original Molecular Descriptors and Multivariate Image Analysis Descriptors

Garkani-Nejad

Ahmadvand

2011

Chromatographia

View full text Add to dashboard Cite

show abstract

“…For these reason in recent years, ANNs have been used to a wide variety of chemical problems such as simulation of mass spectra, ion interaction chromatography, aqueous solubility and partition coefficient, simulation of nuclear magnetic resonance spectra, prediction of bioconcentration factor, solvent effects on reaction rate, prediction of normalized polarity parameter in mixed solvent systems and dissociation constant of acids. [23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39] The main aim of the present work is to develop a QSPR model based on molecular descriptors using ANN for modeling and prediction of E T N values for various solvents (including 216 solvents) with diverse chemical structures. In the first step, a MLR model was constructed.…”

Section: -18mentioning

confidence: 99%

Artificial Neural Network Prediction of Normalized Polarity Parameter for Various Solvents with Diverse Chemical Structures

2007

Bulletin of the Korean Chemical Society

View full text Add to dashboard Cite

Artificial neural networks (ANNs) are successfully developed for the modeling and prediction of normalized polarity parameter (E T N ) of 216 various solvents with diverse chemical structures using a quantitative-structure property relationship. ANN with architecture 5-9-1 is generated using five molecular descriptors appearing in the multi-parameter linear regression (MLR) model. The most positive charge of a hydrogen atom (q + ), total charge in molecule (q t ), molecular volume of solvent (V m ), dipole moment (µ) and polarizability term (π I ) are input descriptors and its output is E T N . It is found that properly selected and trained neural network with 192 solvents could fairly represent the dependence of normalized polarity parameter on molecular descriptors. For evaluation of the predictive power of the generated ANN, an optimized network is applied for prediction of the E T N values of 24 solvents in the prediction set, which are not used in the optimization procedure. Correlation coefficient (R) and root mean square error (RMSE) of 0.903 and 0.0887 for prediction set by MLR model should be compared with the values of 0.985 and 0.0375 by ANN model. These improvements are due to the fact that the E T N of solvents shows non-linear correlations with the molecular descriptors.

show abstract

“…The flexibility of ANN for discovering more complex relationships lead this method to find a wide application in QSAR/QSTR studies, as recently reviewed by Schneider and Wrede 22 . A principal component-artificial neural network (PC-ANN) system, which combines the PCA with ANN, is another PCA-based calibration technique for nonlinear modeling between the PCs and dependent variables 23,24 . The problem of PC selection in PC-ANN is more serious than PCR because of unknown and complex relationship between PCs and dependent variables.…”

Section: Introductionmentioning

confidence: 99%

Application of Multivariate Image Analysis - Quantitative Structure Toxicity Relationship Study for Modeling the Toxicity of Phenol Derivatives

Garkani-Nejad

Salehfard

2013

J. Chil. Chem. Soc.

View full text Add to dashboard Cite

The toxicity of some phenol derivatives over tetrahymena pyriformis were modeled by using multivariate image analysis (MIA) descriptors applied to quantitative structure-toxicity relationship (QSTR) method. MIA descriptors are derived from pixels of two-dimensional (2D) chemical structures that may be built by using any appropriate drawing software. QSTR analysis has been done by using partial least square regression (PLS), correlation ranking-principal component regression (CR-PCR) and correlation ranking-principal component-artificial neural network (CR-PC-ANN) modeling methods. For the ANN models two different weight update functions of Levenberg-Marquardt (LM) and Scaled Conjugate Gradient (SCG) were investigated. Finally, obtained results using different linear and nonlinear modeling methods were compared, showing that PLS and CR-PC-ANN-SCG methods present high performance for prediction of the toxicity of phenol derivatives. Also, these results indicated that MIA descriptors may be useful to predict toxicity of phenol derivatives. Squared correlation coefficient (R 2 ) and standard error using the PLS method for the training set were 0.996 and 0.042 and for the test set were 0.996 and 0.048, respectively. Squared correlation coefficient (R 2 ) and standard error using the CR-PC-ANN-SCG method were respectively, 0.999 and 0.002 for the training set, 0.999 and 0.001 for the test set and 0.999 and 0.002 for the validation set.

show abstract

Structure−Activity Relationship Studies of Carcinogenic Activity of Polycyclic Aromatic Hydrocarbons Using Calculated Molecular Descriptors with Principal Component Analysis and Neural Network Methods

Cited by 62 publications

References 22 publications

Comparative QSRR Modeling of Nitrobenzene Derivatives Based on Original Molecular Descriptors and Multivariate Image Analysis Descriptors

Comparative QSRR Modeling of Nitrobenzene Derivatives Based on Original Molecular Descriptors and Multivariate Image Analysis Descriptors

Artificial Neural Network Prediction of Normalized Polarity Parameter for Various Solvents with Diverse Chemical Structures

Application of Multivariate Image Analysis - Quantitative Structure Toxicity Relationship Study for Modeling the Toxicity of Phenol Derivatives

Contact Info

Product

Resources

About