| Agonists of β2 adrenergic receptor(β2AR)are frequently detected in animal-derived foods,though they have been forbidden to use as growth promoters in animal breeding around the world.However,the current regulation of food safety merely focused on "fire fighting" rather than "fire prevention".It is a challenge to identify the diverse structures and newly synthesized analogues of β2AR agonists which have been proven to be harmful after a long time exposure,and gives rise to serious damages and massive poisoning cases.There is little research about the technique of food safety pre-warning based on molecular level.This technique would give caution for potential poisons and their substitutes and analogues in advance,which not only can eliminate crises of food security in the bud stage,but also significantly reduce the cost of regulation.Besides,there is a shortcoming to the existing immunoassays:this method can only detect one β2AR agonist each time,which causes easy missing defects for diverse β2AR agonists.Therefore the approach of detection of multi-residues needed for rapid high-throughput screening,and it would be promising and urgent in the field of food safety detection.This thesis mainly includes the following content:The first chapter briefly introduces the background of β2AR and their agonists,as well as current analytical methods and their advantages or limitations.Several methods which are applied in this paper such as molecular docking,pharmacophore model and machine learning were also introduced.The safety pre-warning system,which is discussed in the second chapter to the fourth chapter,has three main components,namely molecular docking,pharmacophore model and machine learning.The first part is about the structure-based molecular docking.Ten crystal structures of β2AR selected from Protein Data Bank were used to screen the best structure for docking based classification,and 4LDE was chosen for the model,and the cutoff of the docking score was explored and set to be-8.60 and-8.10.The second part is related to the construction of pharmacophore model.Among three common pharmacophore hypotheses(AAADPR8,AADPR229 and AADPR32)showed a good correlation and predictive coefficients,the regression coefficients of predicted activity versus experimental values for training set compounds are 0.84,0.95 and 0.89,respectively,for the test set,the values are 0.57,0.63 and 0.44,respectively.The third part is about the classification model based on three widely-used machine learning methods(support vector machine,SVM;k nearest neighbor,kNN;random forest,RF).The established models based on selected 12 molecular descriptors reached good accuracy of above 85%for the internal test set and the external test set compounds.In the fifth chapter,we integrated the above mentioned three methods together and successfully developed a comprehensive classification strategy for predictingβ2AR agonists.And more importantly,we developed a free online platform PARA(http://202.127.19.98:8049/)based on the integrated model,PARA could accurately predict all the banned agonists in China and 95% of the agonists collected from various references and databases.Therefore,PARA can be served as a web-based computational warning system of the abuse of β2AR agonists in livestock industry and elsewhere,reducing the risk of potential toxics as food additives.In the last chapter,the molecular modeling was applied to investigate the binding between ligands with receptor,and it can provide information for the design and preparation of multi-residues electrochemical sensor.A type of electrochemical sensor that can simultaneously detect three kinds of β2AR agonists was prepared,and it showed a good linear relationship within the scope of 0.01-100 ng/mL. |
PDF Full Text Request