AI designed, mutation resistant broad neutralizing antibodies against multiple SARS-CoV-2 strains

Area of Science:

• Virology
• Computational Biology
• Immunology

Background:

• The emergence of SARS-CoV-2 variants necessitates rapid development of effective antiviral therapies.
• Existing methods for antibody design are often time-consuming and may not anticipate viral evolution.

Purpose of the Study:

• To develop a comprehensive digital twin for SARS-CoV-2.
• To computationally design and experimentally validate neutralizing antibodies against diverse SARS-CoV-2 strains.
• To assess the potential of AI in predicting viral evolution and guiding therapeutic design.

Main Methods:

• Integration of diverse data types and metadata using machine learning, natural language processing, and protein modeling.
• Development of a digital twin for SARS-CoV-2, focusing on the receptor binding domain (RBD).
• Computational design of antibodies against over 1300 historical strains with 64 mutations.
• Experimental validation of 70 AI-designed antibodies using binding and viral neutralization assays.

Main Results:

• 70 AI-designed antibodies were experimentally validated against multiple SARS-CoV-2 strains, including Omicron variants.
• 14% of designed antibodies showed strong cross-reactivity against the RBD of multiple strains.
• 10 antibodies neutralized the Delta strain (IC50 < 10 µg/ml), and one neutralized Omicron.

Conclusions:

• The developed digital twin approach is effective for computationally designing neutralizing antibodies.
• AI-driven strategies can predict viral evolution and accelerate the development of broad-spectrum antiviral treatments.
• This methodology holds significant promise for future therapeutic interventions against viral pathogens.