QSAR model of phenols generated by deep neural network

Quantitative Structure-Activity Relationships (QSAR) methods are robust techniques to study chemical compounds behavior vis-à-vis biological systems and to assist the drug design process. In this study we establish a QSAR model by means of deep neural networks (DNN) to study the cytotoxicity of phenolic compounds to Tetrahymena pyriformis, an aquatic organism. The established Quantitative Structure-Cytotoxicity Relationship (QSCR) model with DNN shows a high correlation coefficient (0.935) between hydrophobic, electronic and steric descriptors of the studied dataset of phenols and their cytotoxicity. The predictive ability of the network, tested by an external set of phenols, exhibits a correlation coefficient of 0.688.