In the latest breakthrough, Google DeepMind's AlphaFold unveils the intricate molecular dance within life's building blocks, a leap towards deeper biological understanding and drug discovery. While previous versions astounded by predicting 200 million proteins, the new iteration illuminates their collaborations with DNA, RNA, and more. Though limitations exist, its potential to revolutionize drug development garners attention from pharmaceutical giants. Yet, only time and rigorous human studies will reveal AI's true impact on medicine..Sign up for your early morning brew of the BizNews Insider to keep you up to speed with the content that matters. The newsletter will land in your inbox at 5:30am weekdays. Register here..By Lisa Jarvis.Alphabet's artificial intelligence subsidiary, Google DeepMind, has yet again knocked the socks off scientists with its latest iteration of AlphaFold, using the tool to illuminate the intricate dance between some of life's most important molecules. It's an important leap towards a world where technology enables a deeper understanding of human biology and, hopefully, improves our ability to discover new drugs. .___STEADY_PAYWALL___.It was only two years ago that DeepMind's earlier version of AlphaFold blew the scientific community's mind by revealing images of every existing protein on earth — 200 million of them. Those images weren't perfect – they were predictions, not actual snapshots — but even so, they cracked open a raft of possibilities, from speeding basic science to improving the design of new treatments..Read more: Lisa Jarvis: Gene therapy breakthrough suggests imminent cure for hearing loss.But proteins don't exist in isolation. To perform their critical tasks inside the body, they work with a host of partners including other proteins, DNA, RNA and small molecules. Now, AlphaFold is giving scientists a glimpse of those collaborations. It's a major advance over its previous technology as well as over other existing technologies. And although the images again contain imperfections, the tool helpfully ranks its confidence in what it has drawn up..There are some fundamental limits to its utility. One of the most obvious shortcomings is that these are static images, not movies. The biomolecules in our bodies take on various shapes — twisting, stretching, even unwinding — as they interact with their partners in a cell, and AlphaFold is only offering a moment of that motion..But even static images leave plenty of science to unearth. For example, one could know what that interaction looks like for a normal protein, incorporate a mutation into its sequence — much like happens with human disease — and see how the structure has changed, says Fiona Marshall, president of Biomedical Research at Novartis. Novartis is one of two big pharma companies to have forged a drug discovery partnership with Isomorphic Labs, which was launched out of DeepMind in 2021. (The other is Eli Lilly & Co.).It's also worth imagining just how far this technology could go. One of the ultimate goals is to be able to peer inside an entire virtual cell to predict how all the different components interact, Marshall says. Then, scientists could ask what happens when one of those components gets out of whack, much like happens in the course of disease. "This is a step along that trajectory — for me, it's quite a big step," she says..As thrilling as this is, the latest iteration of AlphaFold still only addresses the first step in the yearslong process of getting a medicine to market. Even if computers can eventually come up with a good idea for a drug, that work must be recapitulated in the lab and then tested in humans. While a recent analysis of the nascent field suggests AI-invented drugs outperform human-invented ones in the earliest clinical stages, AI has yet to obviate the need for safety studies or replace clinical trials, the trickiest phase of development for most drugs, when failures far outweigh successes. .Ideally, computers could predict certain properties about a drug, like how it is metabolized and whether it interacts poorly with other medicines or hits unintended targets in the body. The hope is to avoid the unanticipated side effects that can waylay a clinical study..As Max Jaderberg, Isomorphic Labs' chief AI officer, unveiled AlphaFold 3, he noted that the company is building separate tools to get at some of those other steps. Those added tools, alongside this latest version of AlphaFold, could draw in more pharma partners for Isomorphic, as well as advance its own ambitions of inventing new drugs..Although AlphaFold 3 is currently free to academics, the company sees major revenue ahead: Earlier this week, Demis Hassabis, the chief executive officer of both DeepMind and Isomorphic, told Bloomberg Television he plans to "build a multi-hundred billion dollar business — I think it has that potential — as well as be incredibly beneficial for society and humanity.''.The ultimate test for AlphaFold 3, and for all the other ambitious AI efforts in drug development, is whether the machines work faster and more cheaply to come up with medicines that are as good as — and ideally better than — what a human would have invented. Will AI shave years off the long and expensive process of drug discovery? Will it improve success rates in the clinic? Will the drugs suggested by the algorithms be breakthroughs?.Only tried and true human studies and approved medicines will give us the answers. That means we're still some years away from understanding the actual impact of AI on addressing human disease. But given the speed with which the technology is evolving, its increasingly looking like the impact could be vast. .Read also:.Aspen Pharmacare poised to alleviate global obesity drug shortageObesity drug's lingering costs as patients struggle to stop – Weight rebound threatens health and finances🔒 Professionals under pressure: The perils of corporate drug use.© 2024 Bloomberg L.P.