What Have We Learned from Design of Function in Large Proteins?

Area of Science:

• Biochemistry
• Structural Biology
• Computational Biology

Background:

• Designing complex proteins with specific structures and functions computationally is challenging due to their large size and intricate folds.
• Traditional atomistic design methods struggle with large, complex proteins, often leading to misfolding and aggregation issues.

Purpose of the Study:

• To improve the accuracy and scope of computational protein design.
• To enable the design of large, complex proteins with controlled structure and function.

Main Methods:

• Combining evolutionary constraints from natural protein homologs with atomistic computational calculations.
• Utilizing evolutionary constraints to focus design on a reduced, highly relevant sequence space, mitigating misfolding risks.
• Integrating deep learning-based ab initio protein structure prediction with existing design strategies.

Main Results:

• Significantly improved accuracy in designing diverse proteins, including vaccine immunogens, enzymes, and therapeutic proteins.
• Demonstrated ability to optimize proteins that were previously intractable for atomistic design.
• Expanded the potential applicability of computational design to any natural protein with a known sequence.

Conclusions:

• Hybrid approaches combining evolutionary information and atomistic calculations are highly effective for protein design.
• Deep learning advancements enhance the capabilities of computational protein engineering.
• The future of protein engineering will likely be dominated by fully computational methods for discovering and optimizing biomolecular activities.