Identifying glycan motifs using a novel subtree mining approach.

Area of Science:

• Glycobiology and computational biology.

Background:

• Glycans are complex carbohydrates vital for biological processes, including host-pathogen interactions.
• Specific glycan structures, or motifs, dictate the binding specificities of proteins like lectins and antibodies.
• Understanding these motifs is crucial for advancing research into human diseases.

Purpose of the Study:

• To develop an improved method for identifying glycan binding motifs.
• To enhance the characterization of glycan-protein interactions.
• To aid in the interpretation of glycan microarray data.

Main Methods:

• A frequent subtree mining approach was customized by altering glycan notation to include terminal connection information.
• Graph representations of glycans were augmented with nodes indicating linkage presence/absence at specific carbon positions.
• The method was combined with the minimum-redundancy, maximum-relevance (mRMR) feature selection algorithm and trained on glycan microarray data.

Main Results:

• The customized approach successfully identified terminal residues as potential binding motifs.
• When applied to common lectins, the identified motifs aligned with known binding determinants.
• Classifiers trained with these motifs demonstrated strong performance, achieving a median AUC of 0.89 for lectin binding.

Conclusions:

• A novel subtree mining approach, Carbohydrate Classification Accounting for Restricted Linkages (CCARL), was developed for glycan binding classification and motif discovery.
• The CCARL method facilitates the interpretation of glycan microarray experiments.
• This approach will accelerate the discovery of new binding motifs for experimental validation.