Machine Learning Guided Design of High-Affinity ACE2 Decoys for SARS-CoV-2 Neutralization

Area of Science:

• Biochemistry
• Virology
• Computational Biology

Background:

• SARS-CoV-2 infection poses a global health threat.
• Angiotensin-converting enzyme 2 (ACE2) is the host receptor for SARS-CoV-2.
• Engineering soluble ACE2 decoy proteins is a therapeutic strategy to neutralize the virus.

Purpose of the Study:

• To identify novel ACE2 variants with high affinity for the SARS-CoV-2 spike protein using machine learning.
• To reduce the mutational load while maintaining or improving binding affinity compared to existing variants.

Main Methods:

• Utilized a transfer learning algorithm (TLmutation) trained on single ACE2 mutations.
• Applied the model to predict binding affinities of ACE2 double mutants.
• Experimentally validated the binding affinity of promising double mutant variants.

Main Results:

• Identified multiple ACE2 double mutants with enhanced binding affinity to the SARS-CoV-2 spike protein compared to wild-type ACE2.
• Discovered the L79V;N90D double mutant, which exhibits subnanomolar affinity comparable to the previously identified sACE2₂.v.2.4 variant.
• Experimental validation confirmed the predicted high-affinity binding of engineered ACE2 variants.

Conclusions:

• Machine learning approaches, specifically TLmutation, are effective for engineering protein-protein interactions.
• High-affinity ACE2 peptides can be identified for targeting SARS-CoV-2.
• Engineered ACE2 variants represent a promising therapeutic avenue for neutralizing SARS-CoV-2 infection.