Prediction Of Virus-host Protein Interaction And Construction Of Knowledge Base

The virus infects the host tissue through the interaction between the specific proteins on the surface of the virus and the proteins on the surface of the infected cell.When the host tissue is infected,the virus will use the host compound to complete the process of genome replication,protein synthesis,assembly and release of new virus particles,etc.Therefore,it is of great significance to predict the interaction between viral protein and host protein for revealing the mechanism of virus infection.Therefore,how to predict virus-host protein interactions is a major challenge in the field of bioinformatics.There are two main methods to predict the interactions between proteins: biological methods and computational methods.The traditional biological methods are carried out through biological experiments,which is a waste of time.Therefore,computational methods are now being used by more and more researchers to predict the association relationships.Computational methods can make predictions not only using virus-host interaction networks,but also using some biological information about the virus or host.The computational method relying solely on protein-related information is realized by using protein-related information,which ignores the topological structure of protein-related interaction network,and the prediction result is not ideal.Only relying on the virus-host network prediction effect is not very obvious.In order to solve the above problems,the prediction methods based on gene ontology and the prediction methods based on the interaction network(NGRHMDA and NCPHMDA)are studied in this paper.On this basis,by weighting Gaussian kernel similarity and GO term similarity and combining the advantages of NGRHMDA and NCPHMDA algorithms,a virus-host protein interaction prediction model(VHPIPGTNT)was proposed based on GO and network topology.The experimental results showed that: The AUC values and AUPR values of VHPIPGTNT on the zika?virus,HIV,SARS2 and HCV validation sets were all above 0.85.We also compared VHPIPGTNT with NCPHMDA,NGRHMDA,HGIMDA and KATZHMDA.Both the AUC and AUPR values of VHPIPGTNT were higher than those of the contrast method.Finally,although there are many existing virus-host interaction databases,and multiple virus-host interaction databases have been integrated to build VHDB,VHDB lacks coronavirus-host protein interactions,and it also has the problem of unreasonable architecture design.Therefore,in this paper,the virus-host interaction data and related data were sorted out on the basis of VHDB,and the virus-host protein interaction knowledge base VHPPIDB was constructed.VHPPIDB realized the functions of browsing,searching,visualization of virus-host protein interaction networks,submissions and statistics.