CarboGrove: a resource of glycan-binding specificities through analyzed glycan-array datasets from all platforms

Abstract

Glycan arrays continue to be the primary resource for determining the glycan-binding specificity of proteins. The volume and diversity of glycan-array data are increasing, but no common method and resource exist to analyze, integrate, and use the available data. To meet this need, we developed a resource of analyzed glycan-array data called CarboGrove. Using the ability to process and interpret data from any type of glycan array, we populated the database with the results from 35 types of glycan arrays, 13 glycan families, 5 experimental methods, and 19 laboratories or companies. In meta-analyses of glycan-binding proteins, we observed glycan-binding specificities that were not uncovered from single sources. In addition, we confirmed the ability to efficiently optimize selections of glycan-binding proteins to be used in experiments for discriminating between closely related motifs. Through descriptive reports and a programmatically accessible Application Programming Interface, CarboGrove yields unprecedented access to the wealth of glycan-array data being produced and powerful capabilities for both experimentalists and bioinformaticians.

Keywords: binding specificity; database; glycan-binding protein; lectin; microarray.

Fig. 1

Diversity of glycan-array technology and content. A) Several technologies in addition to the planar array are now used to probe glycan arrays. The arrays differ in their modes of glycan or protein presentation and in methods of quantification. B) The sets of glycans contained in the arrays represent diverse types of structures and organisms. C) The rate of development of new arrays has increased since 2016, punctuated by significant advances in glycan-synthesis technology (red text). Monosaccharide symbols follow the SNFG (Symbol Nomenclature for Glycans) system (PMID 26543186, Glycobiology 25: 1323–1324, 2015) details at NCBI.

Fig. 2

Standardized analysis and output. A) MotifFinder analyzes glycan-array data from multiple incubation concentrations, where available, of a given glycan-binding protein. B) The program identifies the family of motifs that represents the specificity of the protein, organized into primary motifs and the subtrees of fine-specificity motifs. C) Among the several visualizations in the output are tabular descriptions and binding curves for each motif. Monosaccharide symbols follow the SNFG (Symbol Nomenclature for Glycans) system (PMID 26543186, Glycobiology 25: 1323–1324, 2015) details at NCBI.

Fig. 3

Breadth of representation of array types of glycan-binding specificities. A) The collection covers a wide variety of glycan families, array providers, and technologies. B) Motif coverage across the arrays. Nearly all motifs are represented on at least 1 array. C) Motif coverage of the glycan-binding proteins.

Fig. 4

Comparison across multiple arrays of results for SNA. A) Reliability score for each of the arrays, where higher scores indicate lower dataset noise. B) CarboGrove reported motifs for each array. Within each array, the motifs are ordered from top to bottom by the relative-binding score given next to the ID and graphical representation of the motif. The monosaccharide symbols follow the SNFG (Symbol Nomenclature for Glycans) system (PMID 26543186, Glycobiology 25: 1323–1324, 2015) details at NCBI.

Fig. 5

Searches for glycan-binding proteins that are specific for selected motifs. A) The motif scores for 118 motifs (listed in the columns) and 785 glycan-binding proteins were hierarchically clustered. The search motifs and top hits are labeled. B) For the top 4 glycan-binding proteins from each search, the proteins’ top motifs and their relative binding scores are indicated. Monosaccharide symbols follow the SNFG (Symbol Nomenclature for Glycans) system (PMID 26543186, Glycobiology 25: 1323–1324, 2015) details at NCBI.

Fig. 6

Searches for glycan-binding proteins that are specific for selected motifs. A) Using models downloaded from CarboGrove that bind any of the comparison motifs, we predicted binding to a series of glycans generated in MotifFinder that contain the motifs, as well as negative-control glycans. B) In the reduced set of 4 lectins (GSL-II, HAA, VVL, and PHA-E), the average binding to the glycans in each group shows a different pattern for each group. The differences in top motifs for the lectins correspond to the differences in binding patterns. Monosaccharide symbols follow the SNFG (Symbol Nomenclature for Glycans) system (PMID 26543186, Glycobiology 25: 1323–1324, 2015) details at NCBI.