Analysis and prediction of mammalian protein glycation

• Published in: Glycobiology (Oxford) Vol. 16; no. 9; pp. 844 - 853
• Main Authors: Johansen, Morten Bo, Kiemer, Lars, Brunak, Søren
• Format: Journal Article
• Language: English
• Published: England Oxford University Press 01.09.2006; Oxford Publishing Limited (England)
• Subjects: 3D structure; advanced glycation end product; AGE; alkaline phosphatase; ALP; Animals; Databases, Protein; Gene Ontology category; glycation; Humans; Internet; lysine; machine learning prediction; Neural Networks (Computer); nonenzymatic glycosylation; post-translational modification; Protein Modification, Translational - genetics; Proteins - genetics; Proteome - genetics; PTM; receiver operating characteristic; ROC; Sequence Analysis, Protein - methods; Software; three-dimensional structure
• DOI: 10.1093/glycob/cwl009
• ISSN: 0959-6658

Glycation is a nonenzymatic process in which proteins react with reducing sugar molecules and thereby impair the function and change the characteristics of the proteins. Glycation is involved in diabetes and aging where the accumulation of glycation products causes side effects. In this study, we statistically investigate the glycation of ε amino groups of lysines and also train a sequence-based predictor. The statistical analysis suggests that acidic amino acids, mainly glutamate, and lysine residues catalyze the glycation of nearby lysines. The catalytic acidic amino acids are found mainly C-terminally from the glycation site, whereas the basic lysine residues are found mainly N-terminally. The predictor was made by combining 60 artificial neural networks in a balloting procedure. The cross-validated Matthews correlation coefficient for the predictor is 0.58, which is quite impressive given the relatively small amount of experimental data available. The method is made available at www.cbs.dtu.dk/services/NetGlycate-1.0.