Research On The Prediction Algorithm Of Kinase-substrate Interactions Based On Hetergeneous Network

Protein phosphorylation is an important post-translational modification and is one of the most basic,common and important modes of regulation in living organisms.It plays many roles in cell metabolism,gene expression and cell signal transduction,etc.Traditional experimental methods for identifying phosphorylation sites are expensive,time-consuming and labor-intensive.With the rapid development of high-throughput technology,a large amount of phosphorylation 'data has been generated.Therefore,the development of efficient computational methods to mine potential data patterns,predict phosphorylation modifications,and provide auxiliary information for biological experiments are conducive to the development of phosphoproteomics.Most of the existing computational methods are based on the local information of amino acid sequences of protein substrate.However,phosphorylation is a complex biological process,not only related to the local sequence of a single substrate,but also often related to other information.In this paper,we use the topology information of the kinase-substrate association network,protein-protein interaction network(PPI),and other similarity networks to effectively predict the unknown kinase-substrates interactions.The main research work of this paper is as follows:Predicting kinase-substrate interactions based on bi-random walk.Firstly,the kinase similarity and the substrate similarity are calculated based on the PPI network,respectively.Then,ClusterONE algorithm is used to cluster the similarity network and the substrate similarity network,respectively.The similarity is adjusted according to the clustering results.Subsequently,the known association data are integrated to construct a kinase-substrate heterogeneous network and the bi-random walk algorithm is used to predict potential kinase-substrate interactions.Compared with other prediction algorithms by 10-fold cross validation,the experimental results demonstrate that this algorithm performs better than other algorithms.In addition,the relevant databases and scientific literature verify the effectiveness of this algorithm for the prediction of kinase-substrate interactions.Predicting kinase-substrate interactions by using matrix completion.Due to the limitations of experimental techniques,only a small portion of the kinase and substrate relationships in existing phosphorylation data are known and there is a large amount of missing information.In order to solve the problem of the absence of kinase-substrate associations,we propose an algorithm for predicting kinase-substrate interactions based on matrix completion.Firstly,the kinase similarity matrix and the substrate similarity matrix are calculated by the local alignment of the amino acid sequences,respectively.Then the original association network is adjusted based on the similarities,and the adjacency matrix of kinase-substrate heterogeneous network is constructed.Finally,the matrix completion is used to fill in the missing information in the adjacency matrix and predict potential kinase-substrate interactions.Ten-fold cross validation is used and the ROC curve is drawn to evaluate the performance of the algorithm.The experimental results show that the algorithm is more accurate and effective in predicting the kinase-substrate interactions than other algorithms.In addition,case study experiments validate the effectiveness of the algorithm in predicting potential kinase-substrate relationships.