GPU-accelerated homology search with MMseqs2

Area of Science:

• Bioinformatics
• Computational Biology
• Structural Biology

Background:

• Protein databases are expanding rapidly, necessitating faster and more sensitive search tools.
• Current computational methods struggle to keep pace with the growing volume of biological data.

Purpose of the Study:

• To introduce and evaluate the performance of the graphics processing unit (GPU)-accelerated MMseqs2 tool.
• To demonstrate the efficiency gains of MMseqs2-GPU for protein sequence and structure analysis.

Main Methods:

• Benchmarking MMseqs2-GPU against CPU-based methods for single-protein searches.
• Evaluating MMseqs2-GPU for large query batch searches using multiple GPUs.
• Assessing the impact of MMseqs2-GPU on protein structure prediction (ColabFold) and structure search (Foldseek) pipelines.

Main Results:

• MMseqs2-GPU achieves 6x faster single-protein searches compared to CPU methods.
• For large batches, MMseqs2-GPU is 2.4x faster than the leading alternative using eight GPUs.
• Accelerates ColabFold by 31.8x and Foldseek by 4-27x.

Conclusions:

• GPU-accelerated MMseqs2 provides a substantial speedup for protein database searching and analysis.
• This tool enhances the efficiency of critical bioinformatics tasks like structure prediction and search.
• MMseqs2-GPU offers a cost-effective and powerful solution for modern biological research.