The explosion in the sophistication and adoption of artificial intelligence has already seen software that can generate text and images based on simple cues, as well as edit videos and find patterns in large datasets.

With experts arguing that the potential of the emerging technology is limitless, several companies are already employing it to seek out new pharmaceuticals that traditional methods have yet to find.

One company based in Fremont, California, wants to go further. It contends to use AI and cutting-edge science at nearly every stage of development, using digital modelling to expedite a usually laborious and expensive process.

For Adityo Prakash, CEO of Verseon, it is not about "trying to cash in on today's buzzword," but rather "to change how the world finds new medicines."

"Of 10,000 human diseases, we can barely treat 500 so far—and even the ones we treat, we treat with three-mile-long side effects lists," he told Newsweek. "What the world needs are fundamentally novel things with new therapeutic profiles. It's not novelty for novelty's sake, because you want better therapeutic profiles, better ways to treat the disease with lower side effects, much greater efficacy."

Traditional drug development relies on forming molecules and testing them to see how they behave—and whether that behavior might counteract a disease—in what Prakash described as "brute force trial and error."

But that means, he estimates, that scientists have so far tested around seven million distinct chemicals, while the number of possible compounds is ten to the power of 33—or a decillion. A 2015 paper put the number of "drug-like" chemicals at ten to the power of 60.

"When you compare seven million to [a] one with 33 zeros, you realize you're not even fishing in a tiny pool by the side of an ocean, you're fishing in a tiny droplet," Prakash said, adding that many molecules in the known conceptual space are often similarly constructed, as pharmaceutical researchers look to already-understood entities when searching for new drugs.

It is in this chasmic, unexplored conceptual space that Verseon hopes to use its own proprietary AI software to find the next medical breakthrough.

The company first builds a 3D model of the protein it wants the drug to target, using AI in the first instance to model how its surface naturally flexes, as well as the interaction of water molecules around it. Through this, it can build up a set of characteristics of the protein's footprint that the new drug will have to bind to.

"Why can't you atom-by-atom design that perfect binder that will fit in the palm [of the protein]?" Prakash asked. "This is how we design just about everything else: every component in the chips in your phone was designed that way. In your computer, the Boeing Dreamliner, the Airbus A380—they were completely designed on the computer."

This does, however, raise a problem. AI machines work by understanding and recombining information, and thrive on large datasets. Verseon's approach requires it to look at molecules that, while organically possible, may never have been synthesized before.

This is where "fundamental advances in quantum physics" come in, Prakash said, and allow for it to model how individual bonds between atoms will form and therefore how the entire molecule they comprise will behave. It also allows individual water molecules to be modeled, to see where they might bind to the protein and alter the shape of its surface.

Rather than cycling through combinations of molecules already discovered and tweaking them to try to fit this pattern, Verseon then uses AI to model the characteristics of potentially entirely undiscovered molecules until it finds one that could be a good fit. Once it does, it looks at molecules similar to that one to see if they might work better.

"Because of this deep quantum modeling, you start finding islands of useful data," Prakash said. "You see that, oh, this particular drug structure is a great binder to this protein. And so is this one, et cetera."

He added that this combination of tools gives "the ability to explore this uncharted chemical space, to find best-in-class, first-in-class drugs that change the standard of care for all for all these major human diseases. That's the true promise of a new, fundamental transformation of 21st-century medicine. That's where it needs to go."

Once the AI has found a number of candidates—which could be composed of radically different things—these are sent to the lab for testing, to see if there are any unintended effects of the molecules that could make them too potent or target proteins too selectively. Prakash illustrated this with a Venn diagram, with those at the center the best potential drug candidates.

While some drug developers might be tempted to stop here, he said, Verseon's AI approach goes further. The company feeds those results back into its AI platform, along with instructions on what characteristics that need improving, and asks it to tweak the molecules to fit the real-world requirements the new drug needs.

The company claims to already be developing 15 candidate drugs for eight major diseases that affect millions worldwide. A potential treatment for reducing strokes and heart attacks is currently in the first phase of clinical trials, which Prakash says has been shown to have a much smaller risk of excessive bleeding than other anticoagulants on the market.

Two candidate oral treatments for eyesight loss associated with diabetes are in the preclinical stage, but Prakash showed retina scans of rats that had been given one of the drugs which show a reversal of hemorrhaging within the eye.

Asked about whether the sort of use of AI Prakash described was within the technology's capabilities or just clever advertising, Oz Alashe, an expert in artificial intelligence and CEO of CybSafe, a data analytics and cyber security company, told Newsweek that while he could not speak to the medical aspects, "using AI to manage large datasets in order to find the thing—that thing being a gap or an anomaly or a pattern, absolutely—that's being done every day, everywhere."

"It doesn't sound like faff marketing" he said. "In principle, can it be done? Can you manage and understand and break down characteristics and that many opportunities using artificial intelligence? 100 percent." However, he could not comment on "what kind of results that's likely to yield or, more importantly, even whether that is really likely to drive efficiencies in the process."

Verseon, first established in 2002, boasts a compelling board of scientific advisors, including Steven Chu, a Nobel Prize winner and former U.S. Energy Secretary, and Trevor M. Jones, a professor under whose stewardship the Wellcome Foundation developed a number of successful medicines, including those for HIV and Hepatitis.

But Prakash and the company's two other founders are mathematical physicists by training, raising the question: why choose to devote two decades to developing a new method of medical research?

He spoke of a keen interest in finding opportunities where different fields are coalescing, in this case biotech and advanced computer science.

"We really felt that there was a big, big societal need in this space," Prakash said. "We wanted to change how the world finds new medicines—because it's obvious, if you live in modern society and rely on modern medicine to keep you living healthier and longer lives, how limited medicine still is [and] how far we still have to go."

About the writer

Aleks Phillips

Aleks Phillips is a Newsweek U.S. News Reporter based in London. His focus is on U.S. politics and the environment. He has covered climate change extensively, as well as healthcare and crime. Aleks joined Newsweek in 2023 from the Daily Express and previously worked for Chemist and Druggist and the Jewish Chronicle. He is a graduate of Cambridge University. Languages: English.

You can get in touch with Aleks by emailing aleks.phillips@newsweek.com.