Hello readers,

Welcome to the AI For All newsletter! Today, we’re talking about the first AI-designed virus given to a patient, and more!

A patient with an E. coli infection that had defeated both antibiotics and every available phage received a treatment designed by a machine. Radical Numerics, a San Francisco lab founded by three members of the team behind the Arc Institute’s Evo models, announced on October 20 that its Omnii-1 model had designed the bacteriophages, which were administered under an FDA-authorized compassionate use pathway. Phage therapy is more than a century old and has had a renaissance as antibiotic resistance spreads, but it depends on finding a virus that happens to attack the right bacterial strain. As Marisa Azad, an infectious disease clinician at the Ottawa Hospital collaborating with the company, put it to the Times, hunting for phages takes a remarkable amount of time.

Omnii-1 skips the hunt. Rather than generating a genome from scratch, it rewrites one region of a known phage: gp17, the tail fiber protein that T7 uses to grab onto a bacterium’s surface. The model was trained on the patient’s resistant strain and told to design legs that could latch onto it. In the lab, 69.9 percent of designed variants with 11 to 15 amino acid changes produced viable phages, against 4.8 percent of random controls, and 2,161 of 6,694 active variants outperformed the wild type. The resulting cocktail cleared the patient’s isolate in a dish before any of it reached a person.

What happens next is unknown, and will be for a while. Phage therapy runs for days or weeks, patients are monitored for months in case the infection returns, and the company won’t discuss the clinical course, citing the risk of identifying the patient. The report is self-published and hasn’t been peer reviewed. Michael Baym, a Harvard microbiologist not involved in the work, told the Times his impulse was skepticism, and that he’d wait to see how novel the viruses turn out to be. CEO Eric Nguyen’s ambition is the part worth watching regardless: compressing phage design from months to, in his words, a matter of days.

Few know it, but AI labs now let you monetize your company’s operational data. And they're spending $15 billion this year on it. 

Your workflows, documentation, processes, and internal convos like Slack are extremely valuable and they're paying anywhere from $100k to $3 million for it per company.

You might think that sounds scary, but all the data is anonymized before, so no PII is sent over. Bezel (Y combinator backed) helps you partner with a lab that wants to license your data specifically if you have more than 20 employees and have existed longer than a year.

Get a free estimate

Read more on IoT For All