Evo-EquiGPS: Synergizing Dynamic Geometry, Global Topology, and Explicit Evolution for High-Precision Enzyme Active...

Area of Science:

• Biochemistry and Structural Biology
• Computational Biology
• Bioinformatics

Background:

• Accurate enzyme active site identification is essential for protein function elucidation and enzyme engineering.
• Next-generation sequencing has led to the discovery of novel enzymes with low sequence similarity, posing challenges for existing computational methods.
• Current methods struggle with static geometric representations, limited graph encoder receptive fields, and evolutionary semantic dilution.

Purpose of the Study:

• To develop a novel computational framework for precise enzyme active site prediction.
• To address the limitations of existing methods in identifying active sites for novel enzymes with low sequence similarity.
• To integrate multidimensional features for enhanced prediction accuracy.

Main Methods:

• Introduction of Evo-EquiGPS, a multimodal graph neural network framework.
• Implementation of a three-branch parallel encoding architecture: dynamic geometric flow, global topological flow, and explicit evolutionary flow.
• Synergistic integration of enzyme sequence semantics, 3D structures, and explicit evolutionary constraints.

Main Results:

• Evo-EquiGPS demonstrated superior performance on datasets with high structural diversity.
• Achieved a significant margin in area under the precision-recall curve (AUPRC) on the TS124 dataset, outperforming SCREEN and GraphEC.
• Showcased strong generalization capabilities on the independent test set CSA112.

Conclusions:

• The Evo-EquiGPS framework significantly enhances the precision of enzyme active site prediction.
• Provides a robust foundation for computational protein functional annotation.
• Advances the field of enzyme engineering and functional genomics.