Study On The Structure-activity Relationship Between The Molecular Structure And Toxicity Of Nitroaromatic Compounds

Nitroaromatic compounds(NACs)are an important type of environmental organic pollutants and have toxic effects such as mutagenic,carcinogenic and teratogenic.Due to limited resources,there are currently a lot of gaps in the toxicity data regarding their potential adverse effects on human health and the environment.With the continuous development of computer technology and artificial intelligence in recent years,structure-activity relationship(SAR)has become one of the principal means to evaluate the harm of environmental pollutants to the human body and the environment.On the one hand,SAR can build models that predict the potential hazards of untested compounds to further fill the gaps in toxicity data,One the other hand,SAR can further clarify the toxicity mechanism of the compounds through the screened descriptors or characteristic substructures.In this study,the SAR method was used to study the relationship between the molecular structures of NACs and mutagenicity and acute toxicity.Quantitative structure activity relationship(QSAR)and classification models were developed using a set of NACs,respectively,which aim to find the molecular descriptors and privileged substructures closely related to mutagenicity and acute toxicity,to further elucidate the mechanism of toxicity of NACs.This thesis mainly contains the following three parts:Firstly,we studied the relationship between the structures of NACs and mutagenicity.In this study,quantitative structure activity relationship(QSAR)and classification models were constructed using a set of NACs based on their mutagenicity against Salmonella typhimurium TA100 strain.For QSAR studies,the best QSAR model contains five simple 2D descriptors with defined physicochemical meaning,in which Q~2_LOO=0.950,R~2=0.967,Q~2_ext=0.836,R~2_ext=0.843,indicating excellent robustness and external prediction performance,greatly exceeded QSAR models constructed without quantum chemical descriptors and models previously reported.For classification studies,seven machine learning methods along with six molecular fingerprints were applied to develop qualitative classification models.Notably,we also obtained some specific molecular properties or privileged substructures responsible for the high mutagenicity of NACs.For the top ten models,their classification accuracy and area under the ROC curve values ranged from 0.828–1.000 and 0.962–1.000,respectively.Notably,we also obtained some specific molecular properties or privileged substructures responsible for the high mutagenicity of NACs.Secondly,we studied the relationship between the structures of NACs and rat oral acute toxicity.In this study,a quantitative structure-toxicity relationship(QSTR)model of acute oral toxicity of NACs to rats was established.The model was assessed by internationally accepted validation metrics and the Organization for economic co-operation and development criteria(OECD).Finally,the best QSTR model contains seven simple 2D descriptors with defined physicochemical meaning,which are Sp Max7?Bh(s),P?VSA?s?1,Eig13?AEA(dm),B06[C-F],B10[N-O],F06[O-O]and F09[C-N].Q~2_LOO=0.7003,R~2=0.7481 of the proposed QSTR model conforms to the statistical verification standard.The model was successfully applied to a real external set prediction and the reliability of the prediction was analyzed and discussed.In addition,rat-mouse and mouse-rat interspecific quantitative toxicity-toxicity relationship(i QTTR)models were established,and the rat-mouse and mouse-rat i QTTR models were validated and toxicity prediction was performed using a true external set of 67 and 265 compounds,respectively.These models have good external predictive power and can be used to rapidly predict the oral acute toxicity of novel or untested NACs in rats within the scope of the model,which is beneficial for environmental risk assessment and regulatory purposes.Thirdly,in the study of acute toxicity classification for NACs in rats,the single-classification models and multi-classification models of acute oral toxicity for NACs in rats were established.The performance of the established classification models was good and 1336 NACs compounds were successfully predicted according to the established multi-classification models,most of which were within the range of AD.For single-classification study,the accuracies of 10-fold cross-validation for the training set were 0.839-0.874 in top ten models,meanwhile,for external test set,the range of accuracy was 0.764-0.824.For multi-classification study,the accuracies of 10-fold cross-validation for the training set were 0.839-0.874 in top ten models,meanwhile,for external test set,the range of accuracy was 0.707-0.776.Furthermore,several privileged substructures for screening NACs with high acute toxicity were identified using the Information gain(IG)and substructure frequency analysis(SFA)method.Moreover,based on the privileged substructures,a representative high-toxic NAC(Parathion)was successfully transformed into low-toxic or non-toxic NACs,which provided a reference for the degradation and transformation of NACs in the environment.Overall,this thesis performs a comprehensively SAR study for the relationship between the structures of NACs and their mutagenicity and acute oral toxicity,numerous predictive models with good performance and the corresponding privileged substructures were obtained.The developed models can be utilized as potential tools for rapidly predicting the mutagenicity and acute oral toxicity of new or untested NACs for environmental hazard assessment and regulatory purposes,and may provide insights into the in vivo toxicity mechanisms of NACs and related compounds.