Researchers from Lawrence Livermore National Laboratory (LLNL), in collaboration with other leading institutions, have successfully used an AI-driven platform to preemptively optimize an antibody to neutralize a broad diversity of SARS-CoV-2 variants.
This pioneering approach, published in the journal Science Advances, represents a significant leap in the fight against rapidly evolving viruses such as SARS-CoV-2, improving future pandemic preparedness and antibody therapy resilience. The paper details the development of 3152-1142, a next-generation antibody derived from AZD3152, a medicine from global biopharmaceutical company AstraZeneca currently approved in Europe and Japan for COVID-19 pre-exposure prophylaxis.
By integrating cutting-edge computational modeling, deep mutational scanning and laboratory validation, scientists have engineered an antibody that restores full potency against multiple potential escape variants, including one that emerged over the course of this work, with the goal of fortifying antibodies against potential future mutations.
"This study is a testament to the power of computational biology and AI in tackling real-world health crises," said Dan Faissol, lead researcher at LLNL. "By integrating machine learning with lab validation, we quickly developed an antibody that countered an emerging threat, proving that we can combat an actively mutating virus."
Addressing the challenge of viral evolution
As the COVID-19 pandemic has demonstrated, SARS-CoV-2 evolves rapidly, rendering many previously effective antibody treatments obsolete. Most clinical antibodies that neutralized early strains lost efficacy against recent Omicron subvariants. AZD3152, developed as a prophylactic treatment for immunocompromised patient populations, also showed susceptibility to viral escape mutations.
To counter this, LLNL and AstraZeneca researchers embarked on a mission to enhance the antibody's effectiveness preemptively. Their approach began with deep mutational scanning, a technique that simulates thousands of possible viral mutations to identify potential weak points in an antibody's binding ability. Scientists discovered that specific mutations at certain positions in the virus's spike protein significantly reduced AZD3152's neutralizing power.
Addressing these vulnerabilities, researchers employed the Generative Unconstrained Intelligent Drug Engineering (GUIDE) computational platform, developed as part of the GUIDE program. The program is executed by the Joint Program Executive Office for Chemical, Biological, Radiological and Nuclear Defense's Joint Project Lead for Chemical, Biological, Radiological, and Nuclear Defense Enabling Biotechnologies, on behalf of the Department of Defense's Chemical and Biological Defense Program. It is aimed at improving biodefense preparedness and cost-effectively discovering medical countermeasure candidates for emerging and unanticipated biothreats.
Researchers used the platform to analyze over 10 billion potential antibody modifications and predict which alterations would enhance binding to SARS-CoV-2 variants, including those not yet in circulation. The top candidates were then tested in the lab to confirm their efficacy.
After two iterative design cycles, the team identified 3152-1142 as the most promising optimized antibody. This new antibody variant demonstrated a 100-fold improvement in potency against a SARS-CoV-2 variant that had previously escaped AZD3152's neutralization.
Implications for future pandemic preparedness
This research builds on previous work by the same team to develop AI-driven antibody optimization as a revolutionary tool in infectious disease management. Unique to this project is the ability to anticipate viral evolution and design therapeutics that remain effective for longer durations, reducing the need for constant redevelopment.
The team envisions to someday have the capability to quickly redesign antibodies for fast approval by the U.S. Food and Drug Administration, similar to how influenza vaccines are approved with an expedited review cycle - the rationale being that researchers are making only a few amino acid changes to a previously rigorously reviewed drug product.
"By looking ahead to address how the virus might evolve, we're not just responding to current threats - we're proactively developing therapeutics to combat potential future viral evolution," said first author Fangqiang Zhu, a computational physicist in LLNL's Biochemical and Biophysical Systems Group.
This project was supported by the Joint Program Executive Office for Chemical, Biological, Radiological and Nuclear Defense's Joint Project Lead for Chemical, Biological, Radiological, and Nuclear Defense Enabling Biotechnologies, in collaboration with the Defense Health Agency COVID funding initiative. Funding also included a grant from the Defense Advanced Research Projects Agency.
For more on GUIDE, visit https://guide.llnl.gov/.
