BridgeNet: a high-efficiency framework integrating sequence and structure for protein and enzyme function prediction.

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Machine Learning in Biology

Background:

• Accurate protein property prediction relies on understanding the complex relationship between protein sequences and their 3D structures.
• Existing methods often struggle to effectively integrate these two distinct data modalities.

Purpose of the Study:

• To develop a novel deep learning framework, BridgeNet, for integrating protein sequence and structural information.
• To enhance protein representation learning for improved downstream prediction tasks.

Main Methods:

• Proposed BridgeNet, a pre-trained deep learning framework utilizing a novel latent environment matrix to align sequence and structural information.
• Employed a modular architecture with sequence encoding, structural encoding, and a bridge module.
• Evaluated the model on enzyme classification, Gene Ontology annotation, coenzyme specificity, and peptide toxicity prediction tasks.

Main Results:

• BridgeNet demonstrated superior performance compared to state-of-the-art models across multiple prediction tasks.
• The framework effectively captures complementary features from sequence and structure without requiring explicit structural inputs during inference.
• Achieved advancements in protein representation learning.

Conclusions:

• BridgeNet offers a scalable and robust solution for integrating diverse protein data modalities.
• The framework significantly advances the field of protein representation learning.
• Enables new applications in computational biology and structural bioinformatics.