Prediction of Glycosylation Sites in Proteins

This chapter introduces ideas of pattern recognition for the prediction of protein glycosylation sites from peptide sequence alone. A general introduction is provided to data-driven prediction methods for solving this problem, including a discussion on artificial neural networks. Also included is a discussion of, and definitions for, the most commonly used performance evaluation measures. Issues concerning data collection are addressed and an explanation is given of how a sequence logo is constructed and can be used to illustrate the sequence specificity around a glycosylation site. There is then a discussion of specific glycosylation linkages for which at least one prediction method has been developed. For each glycosylation type, general information is presented together with proposed sequence motif(s) (if any), a sequence logo, and information on and comparison of existing predictors.