FoldExplorer: Fast and Accurate Protein Structure Search with Sequence-Enhanced Graph Embedding.

Area of Science:

• Structural Biology
• Bioinformatics
• Computational Biology

Background:

• Protein structure databases are rapidly expanding due to accurate prediction methods.
• Efficient and precise protein structure search is increasingly critical.
• Traditional alignment methods struggle with large-scale database searches.

Purpose of the Study:

• To develop a novel deep learning approach for efficient and accurate protein structure search.
• To address the limitations of traditional methods in large-scale database applications.
• To provide a comprehensive view of protein structure space for novel insights.

Main Methods:

• Developed FoldExplorer, a deep learning model utilizing graph attention neural networks and protein language models.
• Jointly encoded structural and sequence information to generate specialized embeddings.
• Evaluated performance on geometric similarity search and classification tasks.

Main Results:

• FoldExplorer demonstrated competitive performance, approaching classical alignment tools.
• Outperformed recent deep learning and sequence-based methods in search and classification.
• Showed effectiveness in searching low-confidence predicted structures.
• Exhibited high efficiency for large-scale database searches.

Conclusions:

• FoldExplorer offers an efficient and accurate solution for protein structure search in large databases.
• The model's embeddings provide a comprehensive view of protein structure space, enabling new insights.
• FoldExplorer is a valuable tool for advancing structural biology and protein universe studies.