Rapid and sensitive protein complex alignment with Foldseek-Multimer

Area of Science:

• Structural bioinformatics
• Computational biology
• Biochemistry

Background:

• Computational structure prediction methods are generating vast amounts of protein complex data.
• Aligning these protein complexes is crucial for deriving biological insights but is computationally intensive.
• Existing alignment methods face scalability challenges with the growing size of structural databases.

Purpose of the Study:

• To develop a computationally efficient method for aligning protein complexes.
• To enable the comparison of billions of protein complex structures within a practical timeframe.
• To facilitate the discovery of novel biological functions and interactions from predicted structures.

Main Methods:

• Foldseek-Multimer identifies compatible chain-to-chain alignments by clustering superposition vectors.
• The method leverages efficient clustering algorithms for rapid identification of related structures.
• Complex alignments are computed based on these pre-identified compatible chain alignments.

Main Results:

• Foldseek-Multimer achieves alignment speeds 3-4 orders of magnitude faster than current gold-standard methods.
• The software can compare billions of protein complex pairs in approximately 11 hours.
• The alignments generated by Foldseek-Multimer are comparable in quality to existing methods.

Conclusions:

• Foldseek-Multimer significantly overcomes the computational bottleneck in protein complex alignment.
• The speed and scalability of Foldseek-Multimer democratize the analysis of large structural datasets.
• This advancement will accelerate discoveries in structural biology and drug development.