Cell: Using AI to prepare for future coronavirus mutations

SARS-CoV-2 is constantly mutating, and each new variant often catches the world off guard

Researchers led by Professor Sai Reddy from the Department of Biosystems Science and Engineering at ETH Zurich basel have now developed a way to use artificial intelligence to answer these questions, possibly even in real time as new variants

Explore the many potential variants

Because viruses are randomly mutated, no one knows exactly how SARS-CoV-2 will evolve in the coming months or years, and which variants will dominate

That's why the new approach takes a holistic approach: For each of these many potential viral variants, it predicts the ability to infect human cells, and whether it will be neutralized by antibodies produced by the immune system that are found in people who have been vaccinated and recovered

Synthetic evolution and machine learning

To establish their method, Reddy and his team used experiments to produce a large number of mutant variants

The spike protein interacts with the ACE2 protein on human cells to cause infection, and antibodies that are vaccinated, infected, or treated with antibodies work

Although the mutant variants analyzed by this group of scientists contain only a fraction of the billions of variants that could theoretically exist — something that would be impossible to test in a lab setting — it does contain a million such variants

By conducting high-throughput experiments and sequencing the DNA of these millions of variants, the researchers determined how these variants successfully interacted

The researchers used the collected data to train a machine learning model that was able to recognize complex patterns and, when given only the DNA sequence of the new variant, could accurately predict whether it would bind to ACE2 to infect and escape neutralizing antibodies

A new generation of antibody therapy

This new approach will help develop next-generation antibody therapies

"Machine learning can help researchers identify which antibodies are likely to be most effective against current and future variants, supporting the development

Identify variants that are able to escape immunity

In addition, the methodology developed by ETH Zurich can be used to support the development

"Of course, no one knows which SARS-CoV-2 variant

Accelerate public health decision-making

Finally, this machine learning approach could also support public health because when new variants emerge, it can quickly predict whether antibodies produced by existing vaccines will be effective

Reddy notes that the technology could also be applied to other viruses that are spreading, such as influenza, because predicting future flu variants may support the development

essay

Deep mutational learning predicts ACE2 binding and antibody escape to combinatorial mutations in the SARS-? CoV-2 receptor binding domain