Why AI can transform pharma — but nature cannot be hurried

{"text":[[{"start":7.45,"text":"Since OpenAI released ChatGPT in 2022, launching a new arms race in the field of generative AI, several industries have been scrambling to figure out how large language model systems could make their workflows more efficient and productive."}],[{"start":21.7,"text":"The world of pharmaceuticals is not immune. The promises of AI in pharma are plentiful and global groups are investing billions of dollars to outcompete one another, while AI-first start-ups are sprouting up in the industry."}],[{"start":35.5,"text":"The pharma sector has long embraced technology to improve the process of discovering and delivering new treatments to patients. “In the research space, AI has been used for decades in the industry,” says Thorsten Rall, global industry lead for life sciences at Capgemini, the IT services group. “Pharma is a data-rich environment and applications like AI for [identifying treatment targets] and molecule design have been around for a while and deployed quite broadly.”"}],[{"start":62.6,"text":"A game-changing moment for use of AI was the release of AlphaFold2 in 2021 by Google DeepMind. AlphaFold, which went on to win the Nobel Prize, predicts a protein’s structure from its amino acid sequence. This “protein-folding” prediction knowledge is crucial to grasping how proteins work and could be essential to understanding previously unknown disease-causing pathogens."}],[{"start":86.8,"text":"“That was a watershed moment,” says Rebecca Paul, vice-president and head of drug design at Isomorphic Labs, the company spun out of Google DeepMind to advance the project. “That was the first time as a scientist I saw a model that could make very accurate predictions about complex biological output and the progress has been steep from there.”"}],[{"start":null,"text":"

"}],[{"start":108.9,"text":"Budget-busting AI bills; AI and trust in air traffic control; chatbots and mental health; factory robots on the march; astronomical implications; rewriting gaming rules; agentic travel agents"}],[{"start":124.15,"text":"The progress has included predictions of how proteins interact with other molecules and DNA, Paul says. AI will be used to design molecules that can target proteins once thought “undruggable”, widening treatment options for patients across many indications, she adds."}],[{"start":139.8,"text":"There is an expectation within the industry that the impact of AI will be enormous. A new report by Capgemini says AI-driven platforms could account for about 60 per cent of new molecular entities within the next decade, up from around 12 per cent of pipelines now."}],[{"start":157.35000000000002,"text":"The way drugs are designed and developed will fundamentally change if the core promises of the new AI era come to pass. The first impact is expected to be in reducing the time it takes to reach the pre-clinical testing of molecules from about four to five years to 12-18 months, according to Rall."}],[{"start":176.05,"text":"The second benefit will be cost savings, since companies will no longer have to run the same level of lab infrastructure, and time will be saved by recruiting patients more quickly and efficiently for clinical trials."}],[{"start":186.55,"text":"Industry experts are also excited about the prospects of AI being able to generate synthetic data good enough to skip the early stages of trials in lab animals. This may allow human trials to begin much sooner, although that appears some way off and is subject to regulatory approval."}],[{"start":204.3,"text":"The rate of failure in drug development has been constant — only one in 10 drugs that begin a clinical trial makes it through the various phases and regulatory approvals. Pharma groups often spend enormous amounts of money on projects that never reach the market."}],[{"start":219.45000000000002,"text":"“If you can use AI . . . in the design phase to come up with molecules that have a better safety profile and reduce the leakage in the funnel, you can have better quality assets,” says Rall."}],[{"start":230.85000000000002,"text":"Danielle Belgrave, vice-president of AI and machine learning at British drugmaker GSK, says the advancement in AI has also been beneficial for “processing data at scale”. With enormous advances in the understanding of human biology, the data generated needs a commensurate level of computing power, which AI has made possible."}],[{"start":252.8,"text":"“It’s more than an AI revolution, it’s been a compute revolution that’s enabled us to scale solutions to understanding human biology,” she says. “The number one thing with AI is [that] being able to process this data at scale, we’re able to do data science in a new way and automate solutions.” This “deep understanding” helps to assess whether a treatment might work, “because medicine isn’t a one-size-fits-all,” she adds."}],[{"start":276.15000000000003,"text":"Since patients with the same diagnosis will often respond to treatments differently, AI can be used to “untangle the [differences] in response and personalise interventions”."}],[{"start":286.75000000000006,"text":"Basecamp Research, a London-based AI life sciences company sequencing new genomes collected from the natural environment, aims to develop personalised medicines from knowledge it generates building its genomic database, says chief executive Glen Gowers. “The scope of drug development massively widens with AI if you can understand biology at a different level, away from the limits of what humans can,” he adds."}],[{"start":311.65000000000003,"text":"Gowers identifies cell and gene therapies — drugs designed to treat diseases at their genetic or cellular source — as a niche where AI could be most useful for drug development."}],[{"start":322.6,"text":"Yet for all the promise of AI, there are questions over how long it will be before a fully AI-developed treatment reaches patients and why it is taking so long."}],[{"start":334.25,"text":"Paul at Isomorphic Labs says it is inevitable that the process will take time “because we cannot compress the physical world and we have to let biology play out”. This means time will still be needed to study the impact and efficacy of a molecule in patients."}],[{"start":350.8,"text":"“Much as we can make incredible predictions and find things we otherwise wouldn’t find, we still have to navigate the physical world — there’s always going to be that time lag,” she says."}],[{"start":361.15000000000003,"text":"“What we’re seeing in the clinic now is a reflection of what AI was doing five to 10 years ago. What we’re doing now, which I believe is a huge step change from what we were doing then, is going to be reflected in five, 10 years’ time.”"}],[{"start":381.40000000000003,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1783090170_1744.mp3"}