The First AI-Designed Vaccine Is Here - And It's Surprisingly Safe
Artificial intelligence is a contentious subject at the best of times. AI models are moving jobs out of the tech industry, and the people who remain need AI for job security. Then, there's all the harm AI data centers are causing, yet every now and then, AI finds a genuinely beneficial niche. Case in point: designing vaccines for the coronavirus. Well, certain breeds of COVID-19, at least. A team of virologists at the University of Cambridge tested a new vaccine and published the results in the Journal of Infection.
The team designed the vaccine to target the Sarbeco family of coronaviruses using an AI known as Digitally Immune Optimised Synthetic Vaccine (DIOSynVax) technology. Long story short, DIOSynVax created a "super antigen" of sorts that was delivered as what a Chemical & Engineering News editorial called a "universal coronavirus vaccine." Oh, and the vaccine was "formulated" into DNA, letting it use "needle-free intradermal administration" technology. Add that to the list of real-life inventions inspired by science fiction (specifically the hypospray from "Star Trek").
The trial consisted of injecting 39 participants, aged 18 to 50, with varying dosages of the vaccine, dubbed pEVAC-PS. Researchers studied the safety, tolerability, immunogenicity (whether it triggered an immune system response), and reactogenicity (whether it produced expected physical responses). According to the findings, all 39 volunteers "tolerated" all the doses without any adverse effects. Minus the expected immunogenicity ones, which consisted of minor COVID-19 episodes, of course.
One test down, many more to go
The pEVAC-PS vaccine trial is noteworthy for being first vaccine designed by an AI that has progressed to human testing. However, science is not built on testing once but rather on repetition. This is especially true during human testing. While researchers have determined that pEVAC-PS is safe, this was only a "phase 1 study" and was subject to several limitations.
For instance, all participants were from the Southampton area in the U.K. (recruited from the NIHR Southampton Clinical Research Facility). Furthermore, it wasn't possible to blind researchers or participants (no, not literally; they just knew which test groups they were in). More importantly, certain groups within the pool of participants hadn't been exposed to specific COVID variants before the trial, which could have unfairly impacted their immune reactions.
According to the research paper, subsequent tests are needed to (and would) shore up several of these issues, especially the final one. After all, pEVAC-PS wouldn't be much of a "universal coronavirus vaccine" if it didn't work on omicron BA.1 or BA.2 variants. Still, researchers are confident that pEVAC-PS could help pave the way for future vaccine technology — mostly because DNA vaccines permit rapid prototyping and are easy to manufacture, and the intradermal administration system cuts down on wasting needles. Between this vaccine and the miracle materials AI has invented, there might be more and more research papers evaluating the efficacy of AI-generated products.