Research On Drug-target Binding Affinity Prediction And Application

Identification of drug target interactions is an important step in drug discovery and research.Good drug target interaction data can provide a strong reference and support for the reuse of existing drugs and the therapeutic targeting analysis of new drugs.Among them,the binding affinity of drugs and targets is a very important kind of interaction data.Compared with the binary classification problem of predicting interaction,the prediction of specific affinity can better reflect the possibility and strength of drug-target binding.It is more efficient and cost-effective to predict affinity with computer methods,which has become a key research in computer-aided drug design.The prediction of drug target binding affinity was studied based on computer technology.At the same time,a drug screening assistant system was designed and developed based on the model.Firstly,the one-dimensional sequence information of drug molecule and target molecule is taken as input,and convolution neural network and long-term and short-term memory network are established to learn the expression of drug and target protein.Finally,the binding fraction is predicted by fully connected neural network.On DAVIS,the mean square error of the model is reduced by 0.002,and the coordination index is increased by 0.006;on the data set KIBA,the mean square error is basically the same,and the coordination index is increased by 0.009,which shows that the model is a good affinity prediction model.Then the experiment was tested to predict the affinity of some drugs and specific target molecules,and the binding affinity ranking was obtained.At the same time,dock docking was carried out for drug target pairs,and their binding free energy scores were calculated.For different drugs,the trend of affinity score and dock docking score of the prediction model are consistent,and some drugs that have been supported by clinical trials and other studies can be screened out,which shows that the model used in this paper has a certain practicality in drug reuse and drug screening.Finally,the development of drug screening assistant system is carried out,and the system is implemented and tested.It enables users to complete the prediction of affinity and basic information retrieval of related drugs on the graphical interface,which is helpful to the preliminary work of drug screening.In conclusion,the performance of the model is good and at the same time,we developed a drug-assisted screening tool system,which aims to provide some help for the preliminary work of drug screening.