Investigation Of Structure-activity Relationship For Organic Toxicity

Structure-activity relationship (SAR) method has been applied widely in medical and environmental chemistry,also had an important role in some fields,such as the designing a potential leading drug, the environmental risk assessment of pollutants, the ecological evaluation and human health evaluation. As a mature means of scientific research, SAR method could build a quantitative mathematical relationship between chemical structure and property/activity to predict the property and activity of compounds. This method was an important approach to make up for the shortage of the experimental data for the parameters of the chemical properties, also, it could effectively reduce the time consuming,laborious and even living creatures in experiment.In this work, SAR method was applied to study the toxicity of 139 various chemicals including saturated monohydric alcohols and monoketones on rats by oral route, and the activity of 139 the histone deacetylases (HDACs) inhibitors. A serial distance matrixes, the conversion of Euclidean matrixes and property extend matrix were proposed in the present work. Based on the data mining with norm index, a serial norm index descriptors were obtained to construct two models for organic toxicity and bioactivity. The accuracy and stability of two models were further evaluated by the leave-one/five/ten-out cross-validation, external validation, mean absolute error and y-randomization test. For model 1 of the acute oral on tats, results demonstrated that it could give satisfactory prediction effect with the high squared relationship coefficient of 0.882; and the high LOO,5-fold, 10-fold cross-validation squared relationship coefficient of 0.833, 0.833 and 0.838;the squared regression coefficients for the training set and the test set of 0.879 and 0.866,respectively. The results demonstrated that the model 1 was robust and accuracy. For model 2 of the histone deacetylases inhibitors, the statistical results suggested that the calculated values were agreed well with experimental values, and the cross-validation results were fit in with the original model results. The relative statistical and validation results showed that the model 2 had a high predictive ability and was stable. Moreover, the application domain of two models were validated to be satisfaction by the leverage approach, and they might be popularized in wide range. The statistical and comparison results showed that this new method could effectively improve the accuracy and predictive ability for predicting the toxicity of chemicals on rats by oral route and the activity of histone deacetylases inhibitors.Finally, using this method, two histone deacetylases inhibitors were designed with highly activity.All the above suggested that these norm index descriptors and this developed formulas could be effectively utilized for the description of structure-activity relationship. Therefore,this norm index-based SAR model could provide an effective method for predicting the toxicity of organic compounds, which also might be valuable for the risk assessment of chemicals and guiding for the leading drug design.