DeepMSA2

DeepMSA2 generates high-quality multiple sequence alignments for protein monomers and complexes by leveraging large-scale genomic and metagenomic datasets to improve the accuracy of protein tertiary and quaternary structure predictions.

Key Features:

• Iterative alignment search: Performs iterative alignment search through genomic and metagenome sequence databases for uniform protein single- and multichain multiple-sequence alignment construction.
• Balanced alignment search: Implements balanced alignment search and effective model selection as core components of the pipeline.
• Metagenomics database integration: Integrates huge metagenomics databases to enhance multiple sequence alignment quality.
• Monomer and complex MSA construction: Generates multiple sequence alignments for both protein monomers and multichain protein complexes.

Scientific Applications:

• Protein tertiary structure prediction: Improves accuracy of protein tertiary structure predictions compared with current state-of-the-art methods.
• Protein quaternary structure prediction: Improves accuracy of protein quaternary structure predictions for protein complexes.
• Deep learning structure prediction enhancement: Provides optimized input information to deep learning-based structure prediction methods including AlphaFold2-Multimer.
• CASP15 participation: Created complex structural models with considerably higher quality than the AlphaFold2-Multimer server (v.2.2.0) in the CASP15 experiment.

Methodology:

DeepMSA2 uses iterative alignment search through genomic and metagenome sequence databases with balanced alignment search and effective model selection.