Isomorphic Labs Moves AI-Designed Drugs to Human Trials

AlphaFold predicts how proteins fold. Whether the same AI lineage can design drugs that actually work in humans is the question Isomorphic Labs is about to put to a clinical test.

The UK biotech, spun out of Google DeepMind in 2021, announced it is preparing to enter human clinical trials with molecules produced by its AI platform. Max Jaderberg, the company's president, confirmed the move on April 16 at WIRED Health in London. "We're gearing up to go into the clinic," he told the audience, according to Wired. "It's going to be a very exciting moment as we go into clinical trials and start seeing the efficacy of these molecules." No specific date was given.

That omission is telling. CEO Demis Hassabis had set a public target of getting AI-designed drugs into human studies by end of 2025. The trials are coming, but later than the company planned.

From protein folding to drug design

Wired reports that Isomorphic's platform is built on AlphaFold, the system that cracked protein structure prediction at scale. Proteins are chains of amino acids that fold into three-dimensional shapes, and those shapes govern function. The number of possible configurations is astronomically large, and researchers had worked on the problem since the 1970s without a general solution. AlphaFold 2, presented by DeepMind in 2020, used deep learning to solve this at a fidelity that transformed structural biology overnight. DeepMind open-sourced the platform a year later, and it rapidly became standard infrastructure for academic research groups worldwide.

AlphaFold 3, released in 2024, pushed further by modeling how proteins interact with small molecules, DNA, RNA, and other ligands rather than treating proteins in isolation. For artificial intelligence in medicine, this represented a shift from descriptive science to a genuine design tool, since drug candidates must bind to specific protein targets in predictable ways. Isomorphic was built to commercialize precisely this capability.

What the trials will actually test

Clinical trials are where computational predictions meet biological reality. A model can estimate binding affinity and selectivity with impressive precision; it cannot anticipate how a molecule behaves inside a patient. Phase I trials assess safety and tolerability before efficacy even enters the picture.

Isomorphic has not disclosed the disease area, the mechanism of action, the specific molecule, or the trial phase. The announcement at WIRED Health was a directional signal, not a data release. The broader artificial intelligence research community has been waiting for exactly this kind of milestone, but the hard evaluation comes when results are published, not when trials are announced.

Context and competing evidence

Isomorphic is not the first to advance an AI-designed drug to human testing. Insilico Medicine, Recursion, and Exscientia have all moved AI-assisted candidates through preclinical work into early clinical stages over the past several years, with uneven results. Exscientia's lead AI-designed oncology compound failed in Phase I in 2023, a concrete reminder that improved design efficiency does not automatically translate into clinical success.

What sets Isomorphic apart is the external validation behind its core science. As Wired notes, AlphaFold's predictions have been independently benchmarked and adopted by thousands of research groups globally. That is a different foundation than proprietary internal models whose real-world accuracy is largely self-reported. Positive clinical signals from Isomorphic would carry real weight across the pharmaceutical industry's ongoing assessment of artificial intelligence in medicine.

The delay past Hassabis's 2025 target is less significant than it might appear. Drug timelines routinely slip for reasons unrelated to the underlying technology, including manufacturing scale-up, formulation work, and regulatory preparation. What matters is that the company is now moving from computational output to biological testing, a transition most AI drug-design ventures announce more often than they actually complete. The clinical data, not the press conference, will determine whether this marks a genuine turning point.

FAQ

What is Isomorphic Labs?
Isomorphic Labs is a UK biotech founded in 2021 as a commercial spinoff from Google DeepMind. It uses AlphaFold's protein structure and molecular interaction modeling to design drug candidates, operating under the Alphabet umbrella alongside its parent research lab.

How does AlphaFold 3 enable drug discovery?
AlphaFold 3 predicts not just protein structures but how proteins interact with small molecules, DNA, and RNA. This is directly applicable to drug design, where a candidate molecule must bind a specific protein target in a predictable way, as detailed by Wired.

Has AI drug design reached human trials before?
Yes. Companies including Insilico Medicine and Exscientia have entered early clinical trials with AI-assisted drug candidates. Results have been mixed: Exscientia's lead compound failed in a 2023 Phase I oncology trial, illustrating that AI-assisted design improves screening odds without guaranteeing efficacy.

Why is the timeline later than originally announced?
Demis Hassabis publicly targeted clinical trials by end of 2025. Isomorphic has not specified the cause of the delay. The company now describes the trials as imminent without committing to a date.