Drug-Target Interactions And Affinity Prediction Resarch

The identification of drug-target relationships(DTRs)is key to drug discovery and other related work,and can provide direction and valuable guidance in areas such as drug discovery,polypharmacology,drug repositioning,drug side effect prediction,etc.DTRs identification tasks can be divided into two categories,drugtarget interactions(DTIs)prediction and drug-target affinity(DTA)prediction,with DTIs prediction is a dichotomous task,while DTA prediction is a regression task.Traditional DTRs identification methods based on biochemical experiments are not only time-consuming and costly,but also have a certain degree of blindness.With the continuous development of deep learning techniques and the increasing power of computer processing,the combination of computer technology with biomedical disciplines has become increasingly close.Computer-aided DTRs identification and prediction methods can alleviate the disadvantages of traditional methods,which are time-consuming,costly and blind,and are therefore favoured and valued by experts and scholars.For DTIs prediction,machine learning based approaches typically perform prediction by extracting information about the chemical structure of the drug,the target and the semantic information in the heterogeneous information network(HIN)formed by the drug-target.Such methods barely take into account the sub-structural information of the HIN and the hidden semantic information between the nodes,so this paper proposes a graph neural network model based on meta-path to predict drug target interactions(GMDTI).Firstly,based on drugs,targets,diseases and side effects in eight datasets,and the eight different types of action relationships between them,the authors construct a drug-target heterogeneous information network(HIN).Then,two different meta-paths are defined to capture the different sub-topology information of HIN and the latent semantic information between different nodes.Especially,the graph neural network method is applied to represent the node by aggregating the information of the first-order neighbor nodes and the nodes of the meta-path.Finally,DTIs prediction is completed effectively by end-to-end learning method.This method takes the first-order topology and the semantic information of meta-path of the drug-target HIN into account,which is helpful to learn more potential drug target relationships.The experiment results show that the proposed method achieves 98.6% in AUC and 94.5% in AUPR,which are higher than all baseline models.At the same time,GMDTI has better robustness than all baseline models by sparsity experiments of datas and reduction experiments of noise.For DTA prediction,the existing methods mainly use the SMILES sequence of drug and the amino acid sequence of the target as input,and then construct some deep learning methods to make predictions.However,these methods still have certain shortcomings,for example,the existing methods only consider one of the SMILES sequence modal information or molecular graph modal information of the drug,resulting in the loss of information and decrease in prediction performance.To address these problems,this paper proposes a drug-target affinity prediction model fusing multimodal information(NMDTA),and verifies that fusing multi-modal information can help improve the accuracy of the DTA prediction task.Experiments on two different datasets are conducted,and the results demonstrates the advancement and effectiveness of the NMDTA method.In summary,this paper studies from two directions of DTIs and DTA,enriches the representation of drugs and targets,and builds two prediction models,which improves the accuracy of DTRs identification tasks and provides new ideas for DTRs identification tasks.