Prediction of Protein-protein Interactions Using Statistical Data Analysis Methods

In this paper, four methods are proposed to predict whether two given proteins interact or not. The four methods applied for prediction of interaction status of proteins are Principle Component Analysis, Multidimensional Scaling, K-means Clustering, and Support Vector Machines. These methods are applied on phylogenetic profiles of Saccharomyces Cerevisiae (baker’s yeast) protein pairs. The phylogenetic profile dataset is constructed from protein sequence information alone by checking the existence of homologs of Saccharomyces Cerevisiae proteins in different species. Several results obtained by applying these methods show that it is possible to make accurate predictions about protein interactions using sequence information alone by the application of statistical data analysis methods. The best classification results are obtained by Support Vector Machines.