PATRICK SCHWAB is not your normal pharmaceutical researcher and his workplace is not your normal pharmaceutical laboratory. It has neither benches nor bubbling liquids. White lab coats are absent, too. Instead, Dr Schwab is dressed entirely in black. But that is fitting attire for one whose workplace is in King’s Cross, an area that was once railway yards and industrial buildings but is now, after a makeover, one of London’s most achingly trendy districts.
Dr Schwab works for GSK, a
PATRICK SCHWAB is not your normal pharmaceutical researcher and his workplace is not your normal pharmaceutical laboratory. It has neither benches nor bubbling liquids. White lab coats are absent, too. Instead, Dr Schwab is dressed entirely in black. But that is fitting attire for one whose workplace is in King’s Cross, an area that was once railway yards and industrial buildings but is now, after a makeover, one of London’s most achingly trendy districts.
Dr Schwab works for GSK, a drug company. His job is to reimagine the future of drugmaking using that similarly trendy branch of computer science, artificial intelligence (AI). He is applying this to transferring as much of the load as possible from glassware to computers: in silico drug design, rather than in vitro.
To this end he is developing a software tool called Phenformer, which he is training to read genomes. By linking genomic information with phenotypes—the biological term for the bodily and behavioural outcomes of particular genetic combinations—Phenformer learns how genes drive disease. That allows it to generate novel hypotheses about illnesses and their underlying mechanisms.
Meet the transformers
Insilico Medicine, a biotech firm in Boston, seems to have been the first to apply the new generation of AI, known as transformer models, to the business of finding drugs. Back in 2019 its researchers wondered whether they could use these to invent new drugs from biological and chemical data. Their first target was idiopathic pulmonary fibrosis, a lung disease.
They began by training an AI on datasets related to this condition and found a promising target protein. A second AI then suggested molecules that would latch onto that protein and change its behaviour, but were not too toxic or unstable. After this human chemists took over, creating and testing the shortlisted molecules. They called the result rentosertib, and it has recently completed successful mid-stage clinical trials. The firm says it took 18 months to arrive at a candidate for development—compared with a more usual timeline of four and a half years.
Insilico now has a pipeline of more than 40 AI-developed drugs it is developing for conditions such as cancer, bowel and kidney disease. And its approach is spreading. One projection suggests annual investment in the field will rise from $3.8bn to $15.2bn between 2025 and 2030.
Tie-ups between pharma firms and AI firms are also becoming common. In 2024 a dozen deals were announced, with a combined value of $10bn according to IQVIA, a health-intelligence company. And last October Eli Lilly, another pharma giant, announced a collaboration with Nvidia, the firm whose chips host the generative neural networks which transformer models rely on, to build the industry’s most powerful supercomputer, and thus speed up drug discovery and development.
Given the pharmaceutical industry’s weird economics—putative drugs entering clinical trials have a 90% failure rate, bringing the cost of developing a successful one to a whopping $2.8bn—even marginal improvements in efficiency would offer big gains. Reports from across the industry suggest that AI has begun to deliver these. It has slashed the preclinical phase (that before human trials begin) to 12-18 months, from three to five years. And it has improved the hit rate. A study published in 2024, of AI-discovered molecules’ performance in early clinical trials, found an 80-90% success rate. This compares with historical averages of 40-65%.
Designing a new drug generally starts by screening small organic molecules for promising biological activity and selecting the likeliest. AI can sift through libraries of tens of billions of molecules, testing properties such as potency, solubility and toxicity using software emulations of those molecules, with no need for real ones to get anywhere near test tubes. Jim Weatherall, one of those in charge of this activity at AstraZeneca, yet another big drug company, says this sorts the wheat from chaff twice as fast as before, and that over 90% of the firm’s small-molecule discovery pipeline is now AI-assisted.
Trial and no error
AI is also helping improve trial design. One approach involves AI “agents” which behave as though they can think and reason. Back at GSK, Kim Branson, head of AI, gave your correspondent a demonstration of an agent-based system called Cogito Forge. When prompted with a question about biology, Cogito Forge is able to write its own code to help answer that question, gather appropriate datasets, glue them together and then create a presentation—complete with charts showing the conclusions it has drawn.
From there it can generate a hypothesis about a disease, including testable predictions, and try to verify or falsify this with a literature search. That search employs three agents: one to look for reasons why the hypothesis is a good one; a second to look for reasons why it isn’t; and a third to judge which of the other two is correct.
Another area where AI shows promise is selecting patients for trials. It can analyse candidates’ health records, biopsies and body scans to identify who might benefit most from a novel drug. Better choice of participants means smaller—and thus faster and cheaper—trials.
The most intriguing use of AI to improve trials, though, is the creation of synthetic patients (sometimes called digital twins) to act as matched controls for real participants. To do this an AI goes through data from past trials and learns to predict what might happen to a participant if they follow the natural course of their condition rather than being treated. Then, when a volunteer is enrolled in a trial and given a drug the AI creates a “patient” with the same characteristics, such as age, weight, existing conditions and disease stage. The drug’s efficacy in the real patient is then measured against the progress of this virtual alternative.
If adopted, the use of synthetic patients would reduce the size of trials’ control arms and could, potentially, eliminate them entirely in some cases. They might also appeal to participants, since their chances of receiving the treatment under test rather than being put into a control group would rise.
Modelling published in 2025 by Unlearn.AI, a digital-twins firm in San Francisco, suggested that this approach could have reduced the size of a control arm in an early Parkinson’s disease trial by 38%, and by 23% in a different study on Alzheimer’s disease. Furthermore, early-stage trials in general, which sometimes lack a control arm, could now introduce these digitally to enhance confidence in early signs of efficacy, and improve how subsequent trials are designed.
Many proteins—molecules increasingly deployed as drugs but which are much larger than conventional drug molecules—have a tendency to jiggle around. That makes determining their precise shapes harder. RNA molecules, the basis of a new class of vaccines, are equally tricksy. And the complex membrane-based structures found in cells’ interiors, more so. But this is an area where understanding is advancing rapidly. AIs are now being trained to model interactions between proteins and other molecules, to predict RNA folding and even to simulate virtual cells.
Recursion, a firm in Salt Lake City, has built an AI “factory” in which millions of human cells are pictured undergoing various chemical and genetic changes. That permits AIs to learn patterns connecting genes and molecular pathways. And Owkin, an AI biotech in New York, is training its model on a vast set of high-resolution molecular data from hospital patients. Tom Clozel, Owkin’s boss, argues that by making discoveries which humans cannot, this work is moving towards true artificial general intelligence in biology.
These departures from the normal tools of pharmaceutical discovery raise the question of whether conventional pharma companies are at risk of disruption. OpenAI, in particular, has been clear about its expectation that models will reach high levels of capability in biology, and is training systems that can reason and make discoveries in the life sciences. For the moment, drug firms have the advantages of a wealth of biological data and the context to understand and use it. At the moment, collaboration is the order of the day. OpenAI, for example, is working with Moderna, a pioneer of RNA vaccines, to accelerate the development of personalised cancer vaccines. But that balance of advantage might change.
Whoever achieves the upper hand, though, if AIs can wring similar efficiencies from clinical trials the probability of a molecule successfully navigating the clinical-trial journey could rise from 5-10% to 9-18%. That may not sound large, but it represents a huge de-risking of the business, with concomitant reductions in the cost of drug development. In the medium term, this could boost investment and the number of drugs arriving on the market. In the longer term—if AIs can solve biology—the technical possibilities for improving human health may be almost limitless.
Curious about the world? To enjoy our mind-expanding science coverage, sign up to Simply Science, our weekly subscriber-only newsletter.
One Subscription.
Get 360° coverage—from daily headlines
to 100 year archives.
Archives
HT App & Website