Research On Protein Function Prediction Method Based On Graph Wavelet

As an important tool for studying proteins,protein-protein interaction(PPI)network helps people to systematically understand the process of biodiversity.Under the trend of scientific and technological development,the maturity of high-throughput sequencing technology has caused an explosive increase in the amount of data in the PPI network,which has promoted the research of protein interaction networks.With the help of computers and mathematical statistics,the data of protein interaction networks is no longer a mess,and it becomes possible to use complex networks to study protein interaction relationships.However,the efficiency of annotating only by experimentally measuring protein functions cannot keep up with the pace of the big data era,and the number of proteins without annotated functions is also increasing with the increase of PPI network data.Therefore,how to efficiently label protein functions has become a key content in biological research.Breaking through the efficiency limitation of experimental prediction and efficiently predicting protein functions through computational methods will lay the foundation for research work in bioinformatics,biomedicine,life mechanism exploration,and broader benefits to human life and health.Graph wavelet has flexible topological structure and can extract the features of network structure graph,therefore it is widely used.Due to the hierarchical,overlapping and modular characteristics of PPI network,this paper uses graph wavelet transform as the research tool to extract PPI network graph features.Furthermore,combined with distance metrics,cluster analysis on protein interaction network is performed to find the modular structure in the protein interaction network,i.e.,protein complex.As the springboard,protein complexes(i.e.,functional modules)are used as auxiliary modules to help predict the proteins with unknown functions contained in protein complexes.From graph wavelets to protein complexs,our analysis finally goes back to function prediction for unannotated proteins.This will open up insight to explore protein function through calculation.