CAZyme3D: a database of 3D structures for carbohydrate-active enzymes

Area of Science:

• Biochemistry and Structural Biology
• Bioinformatics and Computational Biology

Background:

• Carbohydrate Active Enymes (CAZymes) are crucial for biological processes including human health, nutrition, and bioenergy.
• Current methods for CAZyme annotation rely solely on sequence similarity, limiting the discovery of evolutionarily distant but functionally related enzymes.
• A gap exists in dedicated 3D structure databases for CAZymes, hindering structural similarity-based analysis.

Purpose of the Study:

• To develop CAZyme3D, a comprehensive database of 3D structures for CAZymes.
• To establish a novel hierarchical classification system for CAZymes based on structural similarity.
• To provide a tool for structural similarity searches to aid in the discovery of novel CAZymes.

Main Methods:

• Generated 870,740 3D structures using AlphaFold, forming the CAZyme3D Whole dataset.
• Performed structural similarity-based clustering on 188,574 nonredundant CAZyme sequences (ID50 dataset).
• Developed a hierarchical classification including existing CAZy levels and new subclass, structural cluster (SC) group, and SC levels.

Main Results:

• Inter-family clustering grouped CAZy families and clans with similar structural folds into subclasses.
• Intra-family clustering identified structurally similar CAZymes into SCs and SC groups, distinct from sequence-based subfamilies.
• CAZyme3D provides a web-based platform for submitting query sequences or PDB structures for structural similarity searches.

Conclusions:

• CAZyme3D offers a powerful new approach for CAZyme annotation and discovery by leveraging 3D structural information.
• The structural classification system enhances understanding of CAZyme relationships beyond sequence homology.
• This resource is expected to accelerate the identification and characterization of novel CAZymes across diverse research fields.