Using amino acid and peptide composition to predict membrane protein types

Abstract

Membrane proteins play an important role in many biological processes and are attractive drug targets. Determination of membrane protein structures or topologies by experimental methods is expensive and time consuming. Effective computational method in predicting the membrane protein types can provide useful information for large amount of protein sequences emerging in the post-genomic era. Although numerous algorithms have addressed this issue, the methods of extracting efficient protein sequence information are very limit. In this study, we provide a method of extracting high order sequence information with the stepwise discriminant analysis. Some important amino acids and peptides that are distinct for different types of the membrane proteins have been identified and their occurrence frequencies in membrane proteins can be used to predict the types of the membrane proteins. Consequently, an accuracy of 86.5% in the cross-validation test, and 99.8% in the resubstitution test has been achieved for a non-redundant dataset, which includes type-I, type-II, multipass transmembrane proteins, lipid chain-anchored and GPI-anchored membrane proteins. The fingerprint features of the identified peptides in each membrane protein type are also discussed.