Molecular docking study to evaluate the carcinogenic potential and mammalian toxicity of thiophosphonate pesticides by cluster and discriminant analysis

In this paper, the carcinogenic potential and mammalian toxicity on rodents, based on the quantitative relationship models between structure and biological activity (QSAR), were evaluated. The carcinogenicity and acute toxicity were evaluated by docking molecular physicochemical descriptors, on a series of 33 thiophosphonates. These properties, mainly hydrophobicity, electronic distribution, hydrogen bonding characteristics, molecule size and flexibility, and the presence of various pharmacophoric features, influence the behavior of molecule in a living organism, including bioavailability, transport properties, affinity to proteins, reactivity, toxicity, metabolic stability and many others. The model was validated using linear regression methods: principal component analysis (PCA), partial least squares (PLS) and multiple linear regression (MLR); non-linear regression methods: cluster analysis (CA) and discriminant analysis (DA); and neural network analysis: probabilistic neural network (PNN), identifying the best predictor.