AI tool can predict future variants of coronavirus

　　Helps advance the development of next-generation antibody therapeutics and vaccines

　　【Operation "epidemic" in international war】

　　Science and Technology Daily, Beijing, September 7 (Reporter Liu Xia) Swiss scientists have developed a new artificial intelligence (AI) tool that can predict future variants of coronaviruses, including the new coronavirus, and is expected to promote the development of next-generation antibody therapies and vaccines , to provide important reference for the formulation of public health policy.

The related research was published in the latest issue of Cell.

　　To create this new AI tool, the ETH Zurich team generated in the laboratory about 1 million variants of the new coronavirus spike protein, which carry different mutations and combinations of mutations.

The spike protein interacts with the angiotensin-converting enzyme 2 (ACE2) protein on human cells to infect humans, and antibodies obtained from vaccination, infection or antibody therapy work by blocking this mechanism.

Many mutations within the new coronavirus variant occur in this region, which allows the virus to evade the immune system and continue to spread.

　　By performing high-throughput experiments and sequencing, the researchers determined how these variants interact with the ACE2 protein and existing antibody therapies, revealing the extent to which individual potential variants can infect human cells and the extent to which they can evade antibodies.

　　The researchers then used the collected data to train machine-learning models that were able to recognize complex patterns—given just the DNA sequence of a new variant, it could accurately predict whether it would bind to ACE2 to infect and evade neutralization Antibody.

The resulting machine learning model can be used to predict tens of billions of theoretically possible variants, including single and combined mutations, far more than the millions tested in the lab.

　　The researchers say the new approach could help develop next-generation antibody therapies, some of which have already been developed. The method could determine which antibodies have the broadest activity and could also lead to the development of next-generation COVID-19 vaccines.