FALCON@home: a high-throughput protein structure prediction server based on remote homologue recognition

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Protein Structure Prediction

Background:

• Existing protein structure prediction methods struggle with remote homologous proteins, particularly threading approaches.
• Improving fold recognition for distantly related proteins and increasing prediction throughput for proteome-wide studies are significant challenges.

Purpose of the Study:

• To develop FALCON@home, a protein structure prediction server designed for enhanced remote homologue identification.
• To address the limitations of current threading tools by refining the alignment process for proteins with distant evolutionary relationships.

Main Methods:

• FALCON@home extracts conserved regions from structural templates to align query proteins, avoiding issues caused by highly variable regions.
• The system utilizes the Berkeley Open Infrastructure of Network Computing (BOINC) volunteer computing protocol for large-scale computations.
• Leverages computational power from over 20,000 volunteer CPUs for high-throughput processing.

Main Results:

• FALCON@home demonstrates improved remote homologue identification compared to existing threading tools.
• The server achieves a high prediction throughput, processing over 1000 proteins daily.
• FALCON@home-based predictions ranked 12th in the template-based modeling category at CASP11.
• Application to mouse mitochondria proteins revealed correlations between protein half-lives and structural factors.

Conclusions:

• FALCON@home offers a robust solution for identifying remote homologous proteins and increasing prediction throughput.
• The method's focus on conserved regions effectively overcomes limitations in traditional threading approaches.
• The study highlights the utility of volunteer computing for large-scale structural bioinformatics tasks.