New encoding schemes for prediction of protein phosphorylation sites

Abstract

Protein phosphorylation is involved in most cellular functions. Because of the importance of protein phosphorylation, many methods are conducted to identify the phosphorylation sites. Experimental methods for identifying phosphorylation sites are not only costly but also time consuming. Hence, computational methods are highly desired. In this paper, three new encoding methods, BinCTF(Binary-conjoint triad feature), CTF2(new conjoint triad feature) and BinCTF2(Binary-new conjoint triad feature), which are the modification of Binary and CTF encoding, are developed. Then an ensemble support vector machine is applied to predict the phosphorylation sites related to serine (S), threonine (T) and tyrosine (Y) residues. The numerical results indicate that some of the performance of these new methods are better than previous methods.