Methodology Based On Drug-target Protein Interaction

Protein-protein interactions(PPI)is a process in which two or more proteins are combined to produce biochemical effects.Drug-target interaction(DTI)is a process that drugs make an impact by acting on target proteins in living organisms.Therefore,through PPI and DTI and drug-target interactions studies,important implications can be found for drug discovery and drug design.In addition,biological assays cause high experimental and labor costs when studying PPI and DTI.Although researchers made some improvements compared to traditional statistical methods when studying these interactions,there is still much space that can be improved.Due to the fact that hotspot residue data is unbalanced,thus the accuracy of prediction is affected.This thesis proposes a hybrid method,which combines SMOTE and random forest algorithm.The SMOTE algorithm is used to balance the dataset,through which the positive and negative samples reach equilibrium.Thus,the accuracy of the classification can be increased.Through the research on the hotspot residues of PPI interface,it provides theoretical and experimental basis for the study of drug-target protein interaction binding sites.Due to the fact that feature selection can not fully represent the drug or target protein,thus the accuracy of prediction is affected.This thesis proposes a method for the selection of drug and target protein mixed features(MFERF),which combines mixed drug,mixed target feature calculation methods,random forest feature selection and classification method.Two drug feature calculation methods(MACCS and Topology)and target protein feature calculation methods(CTD and Autocorrelation)are used.The calculated drug features and target features are selected separately.The obtained combination features are put into random forests for classification.The experimental results show that the proposed MFE algorithm has higher training efficiency and can obtain higher accuracy in less time.Based on the study of PPI of hotspot residues and drug-target protein interaction,it is easier to further study of the hotspot residues of target proteins which drugs acting on,namely the study of binding sites between drugs and target proteins.