Sequence-Based Prediction of Protein-Carbohydrate Binding Sites Using Support Vector Machines

Area of Science:

• Biochemistry
• Computational Biology
• Bioinformatics

Background:

• Carbohydrate-binding proteins are crucial in various biological processes and diseases, including cancer.
• Accurate identification of carbohydrate-binding sites is essential for understanding protein function and developing targeted therapies.

Purpose of the Study:

• To establish a machine-learning-based method for predicting carbohydrate-binding sites at the residue level.
• To develop a robust and accurate computational tool for identifying protein carbohydrate-binding interfaces.

Main Methods:

• Utilized support vector machines (SVMs) for classification.
• Integrated evolution-derived sequence profiles, sequence information, and predicted solvent accessible surface area.
• Employed 10-fold cross-validation and independent testing for performance evaluation.

Main Results:

• Achieved an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.78 and 0.77 for cross-validation and independent testing, respectively.
• Demonstrated statistically significant enrichment of predicted binding sites in known carbohydrate-binding proteins compared to non-binding proteins.
• Observed a bias of rare alleles towards predicted carbohydrate-binding sites in human mutation data.

Conclusions:

• The developed method, SPRINT-CBH, provides a reasonably accurate and robust prediction of carbohydrate-binding sites.
• The findings highlight the utility of integrating sequence and evolutionary information for predicting protein-carbohydrate interactions.
• SPRINT-CBH is available as an online server, facilitating further research in glycobiology and disease mechanisms.