Automating Immunity

Viruses evolve at a breakneck clip. That keeps medical researchers working overtime to develop vaccines. Now AI is designing inoculations that could help humans stay one step ahead of the bugs.

What’s new: The first AI-generated vaccine has entered clinical trials in the U.S. Nikolai Petrovsky, a professor at Flinders University in South Australia who led the fight against swine flu in 2009, is heading up the project.

How it works: Petrovsky and his colleagues trained a neural network to discover new substances that boost the body’s immune response to a specific trigger. Called the Search Algorithm for Ligands, the model sifted through the medical literature on vaccine efficacy. Then it invented trillions of compounds and predicted their effectiveness. The team synthesized and tested the model’s top suggestions, and the results were promising enough that the U.S. National Institute of Allergy and Infectious Diseases sponsored a year-long trial.

Behind the news: In the war of immunization, the battle line can shift within a single season. This year’s flu vaccine was no match for a late-season mutation, and its effectiveness dropped from 47 percent to 9 percent.

Why it matters: Flu killed an estimated 79,000 people during the 2017-2018 season. AI that fast-tracks effective vaccines could save countless lives.

Takeaway: The number of recent flu victims is a small fraction of the historic toll, and evolution could enable it to come roaring back. AI could allow for rapid response to new strains before they show a glimmer of pandemic potential.