Protein-Protein Interactions Prediction Via Weighted Sparse Representation Based Classification

There are many factors that affect the occurrence of life phenomena in nature,and it is also inseparable from the participation of proteins.The participation of multiple proteins is woven into a network.When performing physiological functions,the expression of proteins is diverse and dynamic.Proteins need to be studied at the holistic,network,and dynamic levels.Protein-protein interactions(PPIs)analysis can help study the pathogenesis of cancer,design new drug targets,and support the development of new drugs.As proteome research enters the era of big data,researchers in bimolecular and related fields have quickly obtained many experimental data.However,using biological experimental methods is time-consuming and costly.In view of this,this paper proposes a novel computational method FCTP-WSRC based on multivariable mutual information.Firstly,we design a new expression method FCTP of protein sequence.FCTP method maps each protein sequence to a feature vector through F-vector,descriptor(C)and descriptor(T).Secondly,principal component analysis(PCA),an effective feature extraction method,is used to extract the most discriminating new feature subset.Finally,we use the weighted sparse representation classifier(WSRC)to make predictions and get good results.The FCTP-WSRC model has accuracies of 96.67%,99.8%and 98.09%for H.pylori,Human and Yeast datasets respectively.In addition,when predicting the significant PPI network CD9,the FCTP-WSRC model performs well and can predict the potential PPI.Therefore,the method proposed in this paper is excellent in performance,easy to implement,and a powerful tool for predicting PPI.