Alignment-based Protein Mutational Landscape Prediction: Doing More with Less.

Area of Science:

• Genomics
• Computational Biology
• Protein Science

Background:

• Genomic data fuels computational prediction of missense variant phenotypic outcomes.
• Accurate prediction methods often rely on multiple sequence alignments (MSAs), which are resource-intensive to generate.
• Protein structure prediction advancements offer faster homology search solutions, like MMseqs2, potentially overcoming MSA bottlenecks.

Purpose of the Study:

• To assess the utility of fast homology search strategies for predicting missense variant outcomes.
• To evaluate the feasibility of generating high-quality, compute-efficient, alignment-based mutational landscape predictions for entire proteomes.

Main Methods:

• Conducted a large-scale assessment of 1.5 million missense variants across 72 protein families.
• Utilized MMseqs2 for fast homology searches to generate necessary sequence alignments.
• Developed and applied a computational pipeline for mutational outcome prediction.

Main Results:

• Demonstrated the feasibility of producing high-quality, alignment-based mutational landscape predictions.
• Showcased the compute-efficiency of the strategy for large-scale proteome analysis.
• Generated the complete human proteome mutational landscape.

Conclusions:

• The MMseqs2-based strategy is effective for efficient and accurate mutational outcome prediction.
• This approach democratizes the generation of alignment-based predictions for large-scale genomic studies.
• The provided human proteome mutational landscape and pipeline offer valuable community resources.