(490g) Reconstruction of Glycosylation Reaction Networks: Integration of Glycomics and Enzyme Data With Computer Models | AIChE

(490g) Reconstruction of Glycosylation Reaction Networks: Integration of Glycomics and Enzyme Data With Computer Models


Liu, G. - Presenter, State University of New York
Neelamegham, S., University at Buffalo, State University of New York

Introduction: Glycosylation is an important post-translational modification that alters more than 50% of mammalian secreted and cell surface proteins [1]. Glycosylated proteins play structural and functional roles in diverse biological processes including inflammation, cancer and development. The recent development of advanced analytical tools, especially mass spectrometry (MS) methods, has heralded the emergence of Glycomics, a field where entire glycomes of cell systems are characterized. These data are starting to be stored in glycosylation specific databases like the Consortium for Functional Glycomics (CFG) Glycan Profiling database (http://www.functionalglycomics.org/glycomics/publicdata/glycoprofiling-n...). It is currently challenging to quantitatively analyze MS data available in such databases in terms of identifying the underlying biosynthetic pathways that result in a particular MS profile. In this regard, since glycosyltransferase enzyme specificity and concentration data are not considered during conventional analysis, a direct linkage between enzyme activity changes to cell phenotype alterations is missing. To address this shortcoming, we developed a computational methodology to reconstruct species-specific and condition-specific cellular glycosylation pathways by integrating glycan structure information from MS data with enzyme-specific information.

Materials and Methods: An application package called GNAT (Glycosylation Network Analysis Toolbox) was developed in order to link MS data to biochemical data available in various databases and in silico glycosylation reaction network models.

Results and Discussion: Three tasks were undertaken. First, glycan information were gathered from MS experiment using GlycoWorkbench [2] to process raw data, annotate structures and calculate distributions. The enzyme-specific information were queried using enzyme databases, including BRENDA [3] and CAZY [4]. The collated data include information on glycan species, kinetic rate constants and enzyme substrate specificity. Second, a network construction algorithm was implemented to build a complete reaction network given the input glycan structures and species-specific enzyme data above. The inference algorithm developed is iterative, with each step using glycosyltransferase or glycosidase reaction mechanisms to either add or remove a monosaccharide residue from a given substrate. The synthesized network that results from the connection of all input MS glycan structures to pre-defined starting material is called the “master pathway”. Third, subset network modeling, hierarchical clustering, principal component analysis and sensitivity analysis were applied to refine the reaction network. In this step, model predictions were compared with MS data to iteratively refine the model. The novel methods described above are implemented in our open-source MATLAB Toolbox, GNAT. The new features described above were tested using two case studies that relate to: i) N-Glycan processing in Chinese hamster ovary (CHO) cell and its mutants [5]; ii) O-glycosylation networks in human promyelocytic leukemia (HL60) cell [6]. These rigorous tests confirm that our approach can be used to construct computer models that quantitatively describe experimentally observed cell-type specific glycosylation reaction networks [7].

Conclusions: GNAT is enriched with new classes and functions to handle MS data, enzyme information, network inference, subset network generation and sensitivity analysis. By integrating glycomics and enzyme database information with modeling, the methods developed enable the synthesis of glycosylation network models that explain the MS data both qualitatively and quantitatively. 


1.Neelamegham, S., et al. Glycobiology, 2011. 21(12): p. 1541-53.

2.Ceroni, A., et al. J Proteome Res, 2008. 7(4): p. 1650-9.

3.Schomburg, I., et al. Nucleic Acids Res, 2013. 41(Database issue): p. D764-72.

4.Park, B.H., et al. Glycobiology, 2010. 20(12): p. 1574-84.

5.North, S.J., et al. J Biol Chem, 2010. 285(8): p. 5759-75.

6.Liu, G., et al. Bioinformatics, 2008. 24(23): p. 2740-7.

7.Liu, G., et al. Bioinformatics, 2013. 29(3): p. 404-6.