Quantitative Structure-Activity Relationships for Toxicity of Phenols Using Regression Analysis and Computational Neural Networks

Abstract

Quantitative structure-toxicity models were developed that directly link the molecular structures of a set of 50 alkylated and/or halogenated phenols with their polar narcosis toxicity, expressed as the negative logarithm of the IGCSO (50Vo growth inhibitory concentration) value in millimoles per liter. Regression analysis and fully connected, feed-forward neural networks were used to develop the models. Two neural network training algorithms (back-propagation and a quasi-Newton method) were employed. The best model was a quasi-Newton neural network that had a root-mean-square error of 0.070 log units for the 45 training set phenols and 0.069 log units for the five cross-validation set phenols.

QSPR Prediction of pKa for Benzoic Acids in Different Solvents

“…The computations were performed with the ADAPT (Automated Data Analysis and Pattern recognition Toolkit) program [37,38], including feature selection routines (genetic algorithm [39] and simulated annealing [40]) and CNN procedures [41]. The CNNs used for this analysis are three-layered, fully connected, feed-forward networks, and they have been described in detail by Jurs and coworkers [41,42].…”

“…The CNNs used for this analysis are three-layered, fully connected, feed-forward networks, and they have been described in detail by Jurs and coworkers [41,42]. The number of neurons of the input layer corresponds to the number of descriptors in the model.…”

Guidance Document on the Validation of (Quantitative) Structure-Activity Relationship [(Q)SAR] Models

“…During supervised learning, the system is forced to assign each object in the training set to a specific class, while during unsupervised learning, the clusters are formed without any prior information. One approach commonly used is multi-layer feed-forward (MLF) networks consisting of three or more layers: one input layer, one output layer and one or more intermediate (hidden) layer (Smiths et al, 1994;Xu et al, 1994;De Saint Laumer et al, 1991).…”

QSPR Prediction of pK for Aliphatic Carboxylic Acids and Anilines in Different Solvents

“…The computations were performed with the ADAPT (Automated Data Analysis and Pattern recognition Toolkit) program [56,57] including feature selection routines (genetic algorithm [58] and simulated annealing [59]) and CNN procedures [60]. The CNNs used are three-layer, fully connected, feed-forward networks, that were employed in our previous papers on the pK a estimation of phenols [4] and benzoic acids [5], and have been described in detail by Jurs and coworkers [60,61].…”