Prediction Of Protein-Protein Interactions Based On Domain

The study of protein-protein interaction(PPI)is of great significance not only for elucidating the regulatory mechanism of life activities but also for disease prevention,diagnosis,and drug design.With the development of life sciences,a large amount of protein domain information has been discovered.In recent years,protein-protein interaction prediction based on domain information has become a hotspot in bioinformatics research.However,most of these methods only considered the quantitative relationship between domains,but do not consider the intrinsic property of domain.To solve this problem,this study proposed a method for predicting protein-protein interactions based on domain's inherent physicochemical property.Based on this method,the predicting of potential nodes of the integrin adhesion network was carried out.At the same time,a proteinprotein interaction prediction model based on the sequence physicochemical property was constructed.The specific research work was summarized as follows:Firstly,based on the physicochemical property of domain,the domaindomain interaction(DDI)prediction was carried out,and the prediction result of DDI was further combined with the domain-domain interaction score in the DDI database to construct a protein-protein interaction prediction model.The positive protein dataset in this model was selected interacting protein pairs in the integrin adhesion network.Through matching on the training set,the optimal parameter combination of the PPI prediction model was obtained.Secondly,based on the results of DDI prediction model,the potential nodes of the integrin adhesion network were predicted.We assumed that the proteinprotein interactions of the interacting domains also existed.The corresponding proteins were matched from the domain interaction results,and proteins were screened using the GO entry information.As a result,178 integrin potential nodes were found.Finally,the sequence physicochemical property was used to build a proteinprotein interaction prediction model.The model combined two feature extraction methods,and the results had high reliability.By comparing the classification results of physicochemical property at the domain level and the sequence level,the results showed that the classification effect of the domain's physicochemical property was significantly higher than those at the sequence level.In general,this study proposed a model for predicting protein-protein interactions based on domain.Based on this method,the GO entries were used to predict potential nodes of the integrin adhesion network.In addition,this study used sequence physicochemical properties to predict protein-protein interactions.This study not only provides guidance for the experimental method to determine the composition of the integrin adhesion network but also provides clues for the study of protein-protein interactions and related disease mechanisms.