AI, new drugs and the appliance of life science

Life sciences is one of Britain’s great success stories. According to Oxford Economics, the sector contributed more than £13 billion to the economy last year, while some of the country’s universities rank among the best in the world for research in the area. It has put the UK in an especially strong position at the intersection of artificial intelligence and life sciences.
The previous government provided £100 million in investment into AI in life sciences last year and Wes Streeting, the new health secretary, has made turning the UK into a “life sciences and medical technology superpower” one of his three key steps to improving the NHS.
While the benefits of AI in other areas of the economy are in their infancy, pharmaceuticals companies have long been in the vanguard of research into machine learning. One of the early AI achievements was the release of AlphaFold in 2018 by DeepMind, a company founded in Britain, which was able to predict protein structures, a scientific problem often compared with the vastly complicated task of mapping the human genome.
Generative AI has already been used to accelerate drug discovery efforts, to provide more efficient clinical trials and even to develop ultra-targeted marketing materials for niche products, with several British businesses at the forefront. So, in the final part of a week-long AI series, we look at some of the country’s most promising start-ups applying the promise of AI to life sciences.
Exscientia is one of a handful of companies employing AI to discover, design and develop new drug candidates, shortening an expensive process that can take four and a half years to between a year and 15 months. It was spun out of the University of Dundee in 2012, having been founded by Andrew Hopkins, its school of life sciences professor.
Rather than chasing success in a specific drug, Exscientia has developed a platform that can be used to research a variety of treatments. It was listed on the Nasdaq stock market in October 2021, raising $510 million in what was the largest initial public offering for a European biotechnology company at that time.
In an uncommon move for a start-up in the life sciences and AI industries, the company was “bootstrapped” in its early years, with no institutional investment. “I think the founders at that point just felt that they didn’t want to go through the traditional route of venture capital funding and heavy dilution [of their equity],” Dave Hallett, 54, the company’s interim chief executive, said.
• Part one: AI and the creative sector
They were also concerned about the time pressures that come with venture money. “They believed, as I do today, that while we’re making significant progress, we’re not going to transform a 150-year-old process in a few years.”
The challenges associated with employing artificial intelligence in the field of drug discovery were not always novel, but they could be more pronounced, he said. “What AI has done is just kind of magnified problems that are already there. If you’re trying to understand really basic biology, but in humans and on a scale, you need to collect a lot of real-world data, but you need to protect the privacy of the humans behind that data.”
BenevolentAI has been one of the more prominent British faces of artificial intelligence in the life sciences industry. The drug discovery start-up went public with a valuation of more than £1 billion in 2022 through a special purpose acquisition company, or Spac. However, since then the company’s share price has fallen by 94 per cent and it has struggled with boardroom strife.
Joerg Moeller, 59, its chief executive, said that at least some of its loss of value could be attributed to the route it took to going public. “Spacs in general haven’t been a success story. If you look at European Spacs, I think they have a share price development that is consistently going down and we haven’t been able to exclude ourselves from that.”
• Part two: AI and business services
Nonetheless, the company can point to real achievements in drug discovery. During the pandemic, it took the start-up only two days to identify baricitinib, a drug originally prescribed to treat rheumatoid arthritis, as a potential treatment for Covid. It soon received emergency approval from regulators in the United States and elsewhere.
Like others in the drug discovery business, it has sought to demystify the black box that often characterises artificial intelligence. “Our focus is on what we are calling explainable AI,” Moeller said. “We want to use it in a very transparent fashion, where we have a track record of what we are doing in terms of the algorithm so that you’re always able to go back and trace back.”
Zoe, a personal nutrition business, is another start-up that has its roots in academia. It was founded in April 2018 by Tim Spector, 66, a professor of genetic epidemiology at King’s College London, along with Jonathan Wolf and George Hadjigeorgiou, both 49, whose backgrounds are in artificial intelligence and consumer apps.
Machine learning is at the core of Zoe’s ability to deliver personalised dietary recommendations to its members. “Nobody eats the same food as anybody else,” Wolf, the company’s chief executive, said. “We need to take your diet as it is and improve it still with things you’re willing to eat and that you like. And so all of those are ways in which you can use AI.”
One of the key challenges facing Zoe and others in the sector is the shortage of data necessary to model the impact of dietary changes on health. Before it launched its first product, the company undertook a three-year study, based on a thousand people, to provide data for its model. “Often that data doesn’t really exist,” Wolf said. “You’ve got to put a lot of work and time and effort to build that data and to make sure it’s accurate and clean and understood.”
Healx is another start-up applying artificial intelligence to drug discovery, although its focus is solely on rare diseases that, under the economics of the traditional process, are unlikely ever to receive funding. It identifies hidden connections between existing chemical compounds and rare diseases, using a myriad of data sources, ranging from biomedical literature to disease datasets.
• Part three: AI and green technology
Tim Guilliams, 40, its co-founder and chief executive, had the idea when he and David Brown, 74, his co-founder, met Nick Sireau. A parent of two children with a rare disease, Sireau was involved in getting a drug approved for their treatment on a budget of less than $10 million, something almost unheard of.
“We met Nick and then thought it’s really an area where we felt AI could make a difference, do it in a different way, reduce the cost, reduce the timelines and make it more scalable,” Guilliams said.
The company has a rare cancer drug that has reached stage two clinical trials after starting the project in 2019, a much shorter timeline than usual.
Guilliams sees the future application of artificial intelligence to drug discovery going well beyond the promise that’s shown at the moment. “Everybody hears about open AI, large-language models, but the opportunity now to apply that level of reasoning and understanding to other languages like biology, I think that’s mind-blowing, though I think that’s going to take a few years.”
Causaly’s goal is to accelerate biomedical research by using AI trained on vast amounts of scientific literature to uncover insights and relationships that might otherwise go unnoticed.
“Understanding disease biology and coming up with these innovations and ideas is a very hard process if you have to read tens of thousands of documents yourself,” Yiannis Kiahopoulos, 43, the company’s co-founder and chief executive, said.
• Part four: AI’s driverless road ahead
Six years ago he and his co-founder sought to use artificial intelligence to create a knowledge graph, illustrating the relationships between concepts across human biology, packaged in software available for scientists both at research institutes and commercial research organisations. From the day Causaly was founded, it teamed up with Novartis, the multinational medicines company,
Causaly says it can read the entire volume of biomedical literature ever published, in seconds, so that researchers can quickly find answers to complex questions.
Two pitfalls of artificial intelligence — its opacity and its tendency to “hallucinate”, or to produce false positives — were particularly significant challenges for the start-up. “If there’s no trust, it means a scientist will not use whatever you are giving her,” Kiahopoulos said. “So it’s a prerequisite for adoption and for creating any kind of benefit in an organisation.”