Researchers in the Netherlands have developed a DNA test for rare diseases that can provide much more comprehensive results than standard diagnostics in a shorter amount of time. A study published today in the New England Journal of Medicine reports that this new approach could replace 15 other genetic tests with a single analysis while increasing the number of patients who successfully receive a diagnosis.
Taken together, all rare diseases affect approximately 400 million people worldwide. Of the more than 7,000 rare diseases that have been identified so far, about 80% are caused by genetic mutations. Obtaining a diagnosis can be critical for people suspected of having a rare disease, offering them perspective for the future, better guidance and treatment, and less uncertainty. However, these patients often have to undergo multiple rounds of testing and wait years before receiving a definitive answer.
The new test is based on long-read genome sequencing, a technology that can read significantly longer stretches of DNA before assembling them into a complete genome. While conventional genomic tests typically read fragments around 300 nucleotides long, long-read sequencing can analyze stretches of up to 20,000 nucleotides at a time. The longer reads make it easier to accurately assemble the full genome, providing a more complete picture of the patient’s genetic makeup.
“Thanks to long reads, we obtain an even more complete view of DNA and can detect complex and hard-to-find abnormalities. We then link these to specific conditions,” says Alexander Hoischen, PhD, professor of genomic technologies at Radboud University Medical Center. “In this way, our knowledge grows and we can make more diagnoses.”
In addition, the test can detect epigenetic modifications in the genome that affect gene function without altering the underlying DNA sequence. Although these modifications can be responsible for some rare disorders, conventional testing methods are currently unable to detect them.
“With current diagnostics, this requires additional specialized tests, but with long reads we capture these modifications as a bonus—two in one,” explains Christian Gilissen, PhD, professor of genome bioinformatics at Radboud University Medical Center.
Earlier this year, the technology was used as part of the National Undiagnosed Hackathon, where over 140 experts across the Netherlands came together to search for a diagnosis for 33 families. Long-read sequencing was used to map their DNA in detail, leading to five new confirmed diagnoses within two days as well as strong suspicion of a diagnosis for another eight families.
As the number of rare disease diagnoses continues to rise, this new test could make the diagnostic process much faster and more efficient. Long-read sequencing could also help researchers identify and investigate the genetic drivers of rare conditions, many of which remain largely understudied.
Based on these findings, Lisenka Vissers, PhD, professor of translational genomics at Radboud University Medical Center, calls for the technology to be adopted worldwide as the first choice diagnostic approach when testing patients suspected to have a rare genetic disorder.
First coined over a decade ago in the aerospace industry to describe a digital replica of a physical object, the concept of a “digital twin” has since found its way into medicine, where it refers to the simulation of a patient’s unique biology. Drawing on multiple layers of patient health data, these computer models promise to predict how a person’s health will evolve over time and how they will respond to any given intervention.
Digital twins represent a transformative shift in medicine, moving from reactive health interventions toward preventive strategies. While this technology is still in early stages, it is already being used to guide personalized cancer treatment, simulate the outcomes of cardiology interventions, and manage complex metabolic diseases like diabetes. However, most applications today are closer to small-scale digital models of a specific tissue or condition rather than a complete digital twin that dynamically adapts to real-world data from each simulated patient.
Credit: Oscar Wong / Getty Images
A convergence of rapid technological advances across multi-omics and artificial intelligence (AI) is priming the development of powerful computational models that can capture intricate biological processes beyond the capabilities of any of their predecessors. As large-scale multi-omics datasets are increasingly combined with clinical and real-time physiological data, digital twins are laying the foundation for a more precise and individualized understanding of human health.
Exploring uncharted territory
Digital twins could have a particularly meaningful impact in areas of medicine where knowledge is limited and currently available technologies have fallen short. One such area is rare diseases. Although rare diseases collectively affect more than 300 million people worldwide, each of the over 7,000 conditions covered under this definition only affects a small number of patients—sometimes even just a single person. This scarcity makes it difficult to study the underlying biology and hinders the development of much-needed treatments and diagnostics.
Ellen M. McDonagh, PhD Group Team Lead European Bioinformatics Institute
“We can use digital twins to address the fact that, with a rare disease, you might only have a handful of patients with that diagnosis,” said Ellen M. McDonagh, PhD, group team lead at the European Bioinformatics Institute (EMBL-EBI) in the U.K. and translational informatics director at Open Targets.
Through a project funded by the Chan Zuckerberg Initiative, McDonagh’s team is developing digital twins of human tissues that combine multi-omics data with a patient’s clinical history and additional phenotype data. Their approach begins by modeling biological processes in healthy tissue, and then bringing in data from common diseases affecting the same tissue to train AI models to predict patterns of dysfunction. This would allow researchers to feed the algorithm data from patients with rare diseases to better understand the underlying biological mechanisms driving each condition.
Integrating diverse layers of multi-omics data will be critical to achieving a more comprehensive understanding of the molecular basis of these rare conditions. In some countries, including the U.K., patients with rare diseases routinely undergo whole-genome or whole-exome sequencing as part of diagnostic testing. However, many of the identified genetic variants remain difficult to interpret with limited current knowledge. By combining genomics with other modalities such as transcriptomics, proteomics, and metabolomics, researchers can develop a more complete picture of the underlying molecular interactions and better determine the relevance of these previously uncharacterized variants.
On this front, a major challenge lies in collecting and integrating data across a wide range of modalities, cohorts, and institutions. To address this, McDonagh’s team is actively developing workflows to standardize data collected from the scientific literature, public datasets, and research environments, enabling more reliable comparisons across datasets and facilitating their integration into digital twin models.
This work also involves efforts to fill gaps in the data, as not all data modalities will be available for every patient. For instance, a computer model could predict what the transcriptomic profile will look like based on genomics data, and vice versa.
“We are benchmarking different methods that can help with predicting missing data, but also evaluating how confident we are in those predictions,” said McDonagh. Knowing which biological processes can be predicted with high confidence, and which cannot, can help researchers draw more robust conclusions and guide future data collection efforts.
As digital twin models keep growing and becoming more refined, they will enable the identification of new therapeutic targets and diagnostic markers, while also forecasting the precise effects an intervention will have on a given person. McDonagh highlights their potential to develop more personalized treatment plans for each patient, adding that, “Monitoring patients over time, one could also predict whether a patient might develop resistance to a given drug and switch them to an alternative treatment.”
Integrating real-time data
Integrating multi-omics data with physiological measurements, obtained from continuous sensors and wearable devices, could help digital twins take a significant step forward in accurately simulating complex and dynamic biological processes. In turn, this could help advance healthcare from a reactive model to a more proactive approach.
Tadao Ooka, MD, PhD Associate Professor University of Yamanashi
“Today, much of medicine begins after a disease has become clinically apparent,” said Tadao Ooka, MD, PhD, associate professor at the University of Yamanashi in Japan. “In contrast, preemptive medicine aims to detect subtle biological changes before symptoms or irreversible damage occur, and to intervene earlier through lifestyle, environmental, pharmacological, or behavioral approaches.”
Achieving such a transformative shift could significantly reduce the burden of chronic diseases such as diabetes, cardiovascular disease, and neurodegenerative disorders. This is becoming an increasingly urgent goal in aging societies, including Japan, where preventing health decline and extending healthy life expectancy are currently major public health priorities.
Ooka’s lab is developing digital twins that integrate patient data from longitudinal multi-omics, wearables, and lifestyle questionnaires. Through Taomics, a company he co-founded, Ooka is also building a platform to collect longitudinal data from patients and healthy individuals. This data is used to create digital twins that can provide users with personalized health recommendations while informing drug discovery and identifying target populations for a more precise approach to clinical development.
“One major objective is to identify biological pathways related to insulin resistance and metabolic dysfunction,” he added. “The goal is not only to predict risk, but also to understand which behaviors or interventions may improve a person’s molecular and metabolic state.”
While multi-omics data can tell researchers what is happening within the body at the molecular level, continuous data obtained from sensors and wearables can provide a deeper insight into what a person is experiencing in daily life, including physical activity, sleep, heart rate, and stress levels.
“The key is to connect these two layers,” said Ooka. “Together, they allow us to move from general advice to personalized, testable, and adaptive recommendations. For example, if a person’s sleep, physical activity, or dietary pattern changes, we can examine how their inflammatory, metabolic, or insulin resistance-related protein signatures change afterward. Conversely, if a molecular pathway appears to be deteriorating, [sensor] data may help identify the behavioral or environmental context behind that change.”
Across all medical specialties, Ooka expects digital twins to make the greatest early impact in diseases where progression is continuous, multifactorial, and strongly influenced by the patient’s lifestyle and environment. These include metabolic diseases, which develop over many years and are shaped by interactions between genetics, environment, and behavioral patterns. Oncology will also be particularly relevant given the complexity of treatment response and resistance processes at the molecular level.
To reach these ambitious goals, however, a number of challenges must be addressed. In addition to ensuring the data used to train digital twin models is robust and reliable, implementation needs to be carefully planned so that digital twins can adapt to and integrate into real-world clinical workflows, reimbursement systems, regulatory frameworks, and ethical governance structures.
“The goal should be to create systems that benefit the broader population,” explained Ooka. “We need to ensure that prediction does not become discrimination, that data is handled securely, and that people receive understandable and actionable recommendations.”
Towards dynamic predictions
In the future, experts expect to see digital twins that integrate multi-omics data with wearable, imaging, clinical, and environmental data to capture the full complexity of human biology, becoming intelligent decision-support platforms. This progress will be underpinned by continued improvements in multi-omics technology, with the coming decade being primed for advances in longitudinal data collection and spatial multi-omics. Coupled with increasingly lower prices, this technology is expected to become much more accessible to researchers and clinicians alike.
Kyung-In Jang, PhD Associate Professor Daegu Gyeongbuk Institute of Science and Technology (DGIST)
“While omics data were once confined to laboratory analysis, emerging wearable technologies now allow real-time detection of certain metabolites and protein markers,” wrote Kyung-In Jang, PhD, associate professor at the Daegu Gyeongbuk Institute of Science and Technology (DGIST) in South Korea. “These innovations support integrating omics into everyday health monitoring, contributing to the accessibility and responsiveness of precision healthcare.”
Within the next decade, McDonagh expects to see the first translational applications of digital twins in the clinic, whether to support diagnosis, patient stratification in clinical trials, or predicting how a patient will respond to a given treatment. “It really does open the door to being able to identify new targets that are causing disease in rare disease patients, but also in more complex, common diseases,” she said. “Ultimately, digital twins will help in the development of new, safer, more effective treatments and more personalized medicine.”
Going forward, Ooka expects medical applications of digital twins to evolve in stages, starting with smaller, disease-specific models, and later becoming large-scale tools that can predict future outcomes and enable patients to alter their disease trajectories through personalized interventions. This evolution will go beyond purely technical improvements, potentially shaking the foundations of healthcare systems as we know them today.
“The field will require new ecosystem models, not only new analytical technologies,” said Ooka. “Medical digital twins cannot be built by academia, industry, hospitals, or technology companies alone. They require long-term participant engagement, trusted data governance, scientific rigor, clinical relevance, and business sustainability.”
Ooka has been actively working on setting up such an ecosystem in Japan through the COI-NEXT initiative, bringing together universities, regional companies, and global partners to return insights derived from their data to local communities.
Credit: Alllex / Getty Images
“Ultimately, I would like to create a system in which individuals can receive personalized health recommendations based on their own longitudinal biological data,” he concluded. “This means moving beyond one-time testing toward a continuous feedback loop: measure, interpret, intervene, and re-measure. At the same time, such a platform could contribute to pharmaceutical research by connecting real-world human biology, lifestyle, and molecular data in a way that supports more precise and efficient drug development.
“My hope is that digital twins will help create a future where healthcare is no longer centered only on diagnosing and treating disease, but on continuously supporting each person’s optimal health throughout life.”
Clara Rodríguez Fernández is a science journalist specializing in biotechnology, medicine, deeptech, and startup innovation. She previously worked as a reporter at Sifted and editor at Labiotech, and she holds an MRes degree in bioengineering from Imperial College London.
Despite an estimated 190 million women and girls around the world living with endometriosis, a chronic and painful gynecological condition, no disease-modifying therapy has yet been approved to treat it. Léa Wenger, PhD, and her colleagues at Cyclana Bio are aiming to fix this.
Endometriosis occurs when endometrial tissue grows outside the uterus, causing inflammation, pain, and sometimes scarring and fertility problems. Although this condition was historically neglected in terms of research and development, Cyclana is now one of a small but growing group of companies trying to develop more effective endometriosis treatments. After completing a veterinary degree, Wenger shifted away from clinical practice when she discovered a passion for biomedical research during her PhD at the University of Cambridge. During this time, Wenger was diagnosed with endometriosis, which led her to co-found Cyclana Bio in 2024 with Kevin Chalut, PhD, who was her colleague at Altos Labs at the time.
The company joined the Babraham accelerator program last year and has already raised an oversubscribed £5 million ($6.8 million) pre-seed round. Wenger spoke to Inside Precision Medicine’s senior editor, Helen Albert, about her inspirations, career, and what she and her colleagues are hoping to achieve at Cyclana.
Q: What inspired you to become a scientist?
Léa Wenger, PhD
Léa Wenger: I was always very curious as a child. What drove me directly to science, rather than going for any other subjects, was my desire to be a vet. I wanted to be a vet and knew vets needed to know about science, so I decided to learn all the science I could. The irony of that was that I didn’t end up practicing a single day of veterinary medicine, but it got me into the doors of institutions where they teach you veterinary medicine in a way that was very scientific and research-driven. I really discovered a passion for science at that point, a passion for actually understanding things that we don’t know. I was exposed to this idea of driving knowledge where it isn’t present, and that was really what got me excited about research. That’s when I effectively shifted from the veterinary medicine career to the more traditional biomedical research route.
Q: What made you decide to go into biotech rather than staying in academia?
Wenger: I think the frustration I had in academia was that the system was set up to do a one-person, one-project type of research. That can be fun in some ways, but for me, it didn’t really address impact in the way that I really wanted it to. I wanted to feel like I was working towards creating discovery, translating it, and being able to improve patient lives. I just felt that biotech was a better conduit for that because it was based on faster-moving collaborative teamwork.
I was working in neurodegeneration at the time and on organoid models made of 3D stem cell-derived complex architectures. Organoid models are incredibly good at reproducing human development. But when you’re looking at neurodegenerative diseases that happen with age, it’s a lot harder. Aging in a dish is really hard to reproduce.
Cyclana Bio scientists working in the LiveLabs laboratory, from left to right, Kevin Chalut, PhD, cso and co-founder, Léa Wenger, PhD, ceo and co-founder, Siiri Salooma, PhD, founding scientist, and Tom Wyatt, PhD, founding scientist.
It was at exactly then that I wanted to go down this research route in more detail that Altos Labs opened in Cambridge. The company had a thesis of “Let’s try and do real discovery science, deep, groundbreaking science,” but in a biotech environment where you’re much more collaborative. That really attracted me at the time, and so I applied to work there after my PhD. Luckily enough, they took a chance on me and believed in me.
I loved it and I learned a huge amount. Not just in terms of how you build discovery programs from the ground up, but also how you work in a team, how you focus, and how you align incentives in biotech. I think it completely shifted my mindset away from simple academic curiosity to, “How do we drive that curiosity towards impact as quickly as possible?” The bar, in my opinion, is somewhat higher than in academia because you’re not just saying, “Is this good enough to publish?” You’re saying, “Is this a therapy? Is this actually good enough to put into a human and help them and not harm them?”
Q: What made you decide to found Cyclana Bio?
Wenger: I was in an epigenetics lab within Altos, and my co-founder was actually one of the group leaders there working on the extracellular matrix. The more I worked with him, the more I realized how massively important it is in guiding how cells behave. You can get completely different responses from a cell depending on what environment it’s in. I got really interested in that interface between the epigenetics, the gene level regulation, and the [extracellular] matrix.
I was doing a lot of discovery science there, but during that time, I also developed endometriosis. I was in my mid-to-late 20s when my symptoms started, and they got worse very quickly. Like every scientist who gets diagnosed with a condition, I nerded out on the disease. In my spare time, I downloaded all the data and looked into what research was there, and very quickly realized that there wasn’t much information available.
There’s not much that we know about the disease and how it happens. People are still debating the causes and drivers of endometriosis. That was interesting and another area of unknown, which has always been what I was attracted to. I’d always been passionate about women’s health, but never really had the opportunity to do something about it.
There’s an easy, non-invasive way of getting access to cells to study endometriosis because menstrual fluid is built and shed every month from your endometrium. It’s built and shed in healthy women, in women with endometriosis, and in women with other conditions. On top of that, biopsies are actually way more common in gynecology than in a lot of other conditions. So I realized this was a huge opportunity to do this tissue-level discovery that we were so passionate about, but for a cause that I really believed in, in a field that was unknown.
I spoke to the CEO at Altos at the time and explained what I wanted to do. He was supportive and thought it was an interesting idea, but ultimately, the indication didn’t align with the priorities of Altos—of looking into age-related diseases. So that’s when we left and started Cyclana Bio.
Q: How easy was it to start the company?
Wenger: We were very lucky in that we got into the Babraham accelerator program very quickly, last May. That was important because not only did it give us validation that someone had actually picked us and said this is a good idea, but it also gave us lab space and access.
We did take quite a bit of a risk. Both my co-founder and I left without having raised funding or grants to start the company. For a short while, we were living off our savings and also paying for some very preliminary science and our first scientist to try and get some data going. That was in May [2025], but quite quickly, we got a bit of traction. By July, we had our first investment term sheet because we started fundraising immediately. Then by September, we were oversubscribed. We finalized the closing in October–November for a £5 million pre-seed round.
I think along the way, we basically just had to assume it was going to happen. We were building the company as if we had the money already, although we were always very open with Tom [Wyatt], the scientist who joined us first, about how much funding we had when he joined.
We’re nine people now and have some amazing scientists who have joined the team, including a great CTO who was also at Altos beforehand. We are still growing as we speak. It’s funny how science brings so much more science.
Q: What are you trying to achieve at Cyclana?
Wenger: Our main aim is to get at least one therapy that’s truly disease-reversing to the clinic. Based on a lot of research, we know that the extracellular matrix can guide how cells respond. It can act as a sink for particular signal factors. Sometimes it can sequester or deliver things like growth factors or inflammatory signals, but it also massively changes how the cell is interacting with its neighbors.
It’s a key component of a positive runaway effect that happens in lots of chronic inflammatory diseases and in some cancers. A lot of the time, when trying to develop treatments, we focus on the cells and whether we can stop that inflammation. What we think is, if you don’t address problems with the matrix, you are not going to cure the disease. You’re effectively just going to mitigate the side effects, and this matrix is going to act like a memory of the disease. This means that if you stop the treatment, it comes back because the matrix issue hasn’t been solved.
We think that that’s a big element of what’s going wrong with endometriosis. Lesions are removed surgically and then they come back. We really think the diseased extracellular matrix is very much driving the pro-inflammatory phenotype, and that if we don’t address that, we don’t actually get to the point where we are curing the disease. We want to see if we can effectively reverse the phenotypes and if we can effectively get to a cure by stopping this recurrent feedback loop.
We haven’t settled on an exact target or modality yet. We’re exploring a few different targets, and I think based on exactly what mechanism we want to go after, we will determine what the best modality is. We want to be sure about the science, very sure about the target, and then make that target work.
Q: Where does precision medicine come into your strategy?
Wenger: Our overall strategy is based on how we see endometriosis as a whole, but I think endometriosis hopefully won’t be viewed like that much longer. We hope that there’s going to be much better stratification and classification of the disease, because it manifests very differently in different women.
Although we think the extracellular matrix might be a common mechanism, we’re building a research platform where we will hopefully find out for sure because we’re collecting data. We’ve got an ongoing observational clinical study where we’re collecting biopsy tissue, menstrual fluids, clinical data, and blood from women, either with or without endometriosis.
We’re collecting that data, looking at the tissue, the proteins, and the architecture, but also isolating cells to test in our models. Then, when we start perturbing with particular interventions that we think might reverse the disease’s impact on the matrix or have different effects on the cells, we might start seeing patterns as to which types of women with endometriosis respond well to different treatments.
It’s going pretty well so far, thanks to our clinical collaborators and participants who have donated samples. Menstrual fluid is a very good way of getting samples from seemingly healthy women, because they don’t need to go to the doctor, [they] just send us a sample. We are also collecting tissue biopsies during routine gynecological procedures to minimize invasiveness and inconvenience.
So far, there’s lots of variability, which was what we expected and which is why we want to collect [samples from] a high number of donors. Not because we think that variability is noise, but because variability is signal. It can tell us more about the nuances of the disease in these different manifestations.
Q: What has the experience of being a CEO and biotech founder been like so far?
Wenger: It’s definitely been a steep learning curve. I think that’s also why it’s been so fulfilling, because I do love being in an environment where I’m not complacent, where I’m always learning.
To some extent, because we had so much freedom at Altos to drive our own projects, I had exposure already to the pure project management side of science, so that didn’t seem quite as much of a step up.
Obviously, there’s a huge business, commercial, and legal dimension that I never had thought about before. But I have been trying to learn as much as I can, as quickly as I can, from others. One thing that the biotech field is quite good at is volunteering information. You go to any sort of networking event, or you meet someone from the industry, and they are often very willing to talk to you about what they’re interested in, but also about what you’re doing, and give any advice they might have.
I’ve met many people who have helped me along the way and who have shared their opinions with me. I walked in expecting academia to be way more collaborative than the biotech industry, but actually, I’ve been very pleasantly surprised with my experience.
Q: Can you share any key learning experiences from the last year?
Wenger: If you have scientific training or you can think in a scientific way, going into the field of business or building a company is somewhat similar. There’s lots of information and lots of alternative paths that you can take, just like in scientific discovery, and there is differently weighted evidence as to which paths are the best ones to take. Once you have a certain amount of information, you can then take the best educated guess. That’s how I’ve gone about building the company. For example, when I started, I was told by a friend, “If you’re starting a biotech, you’re going to need to raise venture capital.” They gave me a book called Venture Deals, which is a very good book that explains how funds work. I read that book and felt I was better equipped to talk to people at the fundraisers. I think the first thing I’d say is, when going into any sort of field, try to really understand how and why that field exists and what are the structures that define its environment. Then you can put context into how people work. As a first-time founder, you might think, “I’m going to find investors, and if they believe in me, they’ll invest.” But there’s so much more to running a venture capital fund. Those things are important to know when framing your discussion.
Something I would do differently is not do everything at once. I left my job, started the company, started the science, started building the network, and started fundraising at the same time. There was always this pressure when I was meeting people that I also had to get them to invest. I think looking back on it, if I could start over, I probably would have spent a few more months trying to build my network and understanding the field better before I started having those investment conversations. It still worked out for us, we still raised funds, but it was stressful. Networking events were very high stakes!
Q: How has the endometriosis space changed in recent years?
Wenger: Gedeon Richter purchasing the Celmatix portfolio and backing FimmCyte are very good signs that people are trying again. I think endometriosis has been plagued by failures in clinical trials, and I think now we’re finally seeing some non-hormonal options being tested, which makes me hopeful. Some will fail, some will succeed, and the successes will drive more interest and availability of funding and hopefully, more successes in the future. I’m really looking forward to seeing the results from some of those clinical trials because I think the more solutions we have for women, the better.
Q: Is funding in the overall field of women’s health changing for the better?
Wenger: Absolutely. I think the funding environment is more open to women’s health. I think that’s been helped by the World Economic Forum and McKinsey Health Institute driving the message of value there. There is excitement, I think, and more funding, especially privately.
I’m not sure about public funding. I do think that on the public funding side, we have a trend of saying, “We should fund women’s health, let’s look for quick wins.” I think that’s a bit of a problem with any field that’s been somewhat left behind, once we realize that we need to bring it back. The risk will be funding the wrong things or putting too much of the money into solutions that may not be revolutionary because they don’t have the foundational science to back them up.
It can also be easy to get stuck in the valley of death between seed and late-stage funding. But I do think that there are incredible scientists moving into the field, and there are some great companies starting up. So even if there is this bias towards pre-seed or seed funding, you only need a couple of those companies to have some really promising data, and they will be funded. The bar might be higher than in other fields, but if you produce groundbreaking discoveries, there will be money.
Q: What advice would you give other new founders starting to build their companies?
Wenger: Just follow your gut and your dream. That’s the most important thing. I started Cyclana because I thought this needed to happen and we needed to look into endometriosis. I thought it was a bit hypocritical of me to think we needed to do something and not do it, despite having the training and the skills to try and find a solution. If you really believe something needs to happen in the world, startups are the best way to feel like you are driving that change and contributing to seeing the change that you want to happen. Whether it succeeds or not, you won’t wake up thinking, “What am I doing this for?” You’ll just be thinking, “I really hope that we don’t fail!”
Helen Albert is senior editor at Inside Precision Medicine and a freelance science journalist. Prior to going freelance, she was editor-in-chief at Labiotech, an English-language, digital publication based in Berlin focusing on the European biotech industry. Before moving to Germany, she worked at a range of different science and health-focused publications in London. She was editor of The Biochemist magazine and blog, but also worked as a senior reporter at Springer Nature’s medwireNews for a number of years, as well as freelancing for various international publications. She has written for New Scientist, Chemistry World, Biodesigned, The BMJ, Forbes, Science Business, Cosmos magazine, and GEN. Helen has academic degrees in genetics and anthropology, and also spent some time early in her career working at the Sanger Institute in Cambridge before deciding to move into journalism.
Google DeepMind is funding research into the potential dangers of situations where millions of different AI agents interact with each other online.
According to Rohin Shah, who directs the company’s AGI safety and alignment research, the mass-market arrival of agents that can carry out tasks without human oversight and follow instructions given to them by other agents creates a whole new class of risk.
In an effort to address this, Google DeepMind—which made agent-based tools a centerpiece of Google I/O last month—has teamed up with several other organizations to announce a $10 million funding pot for researchers to study the behavior of multi-agent systems and come up with ways to prevent unsafe scenarios. Joining Google DeepMind are Schmidt Sciences, a philanthropic foundation set up by Eric and Wendy Schmidt; ARIA, the UK government’s moonshot agency; the Cooperative AI foundation, a UK-based nonprofit research outfit; and Google’s charitable arm, Google.org.
I asked Shah and James Fox, who leads the Science of Trustworthy AI program at Schmidt Sciences, what they hope to achieve with that $10 million. It’s no small sum, but it’s dwarfed by the budgets commanded by Google DeepMind’s own research teams.
The aim is to kick-start research outside tech companies, says Shah: “The strength of academia is that it can look really quite far into the future and do the kind of work that isn’t top of mind at industry labs.”
“The main issue is that there just isn’t really a field of research for multi-agent safety yet,” he adds. “And we would like there to be.”
The concern is that as more and more AI agents get deployed and begin working together, we could hit a tipping point where imagined scenarios become real. “We see this with humanity, too,” says Shah. “Our institutions can accomplish things that no individual human can.”
Shah thinks that we have a few more months to go before agents are deployed throughout the economy in numbers that make potential risks a real concern. He wants to get ahead of that moment.
Risky business
What risks are we talking about, exactly? The possibilities that Shah and Fox have in mind mostly boil down to supercharged versions of bad things that happen on the internet already: scams, prompt injections (where an AI agent is fed malicious instructions, turning it into a self-guiding piece of malware), other forms of cyberattack. We look at what humans do now and ask what the agent version of that would be, says Shah.
“We’ve got this digital commons that is integral to how society works, and you really want to ensure that this doesn’t descend into just absolute anarchy,” says Fox.
(I asked Shah if they were considering any worst-case scenarios more on the doomer end of the spectrum, such as widespread economic collapse. “Certainly not if we’re talking by the end of the year,” he said. That’s only six months away! He laughed. “Okay, a while after that.”)
Shah and Fox both think that the only way to understand what might happen when large numbers of multi-agent systems interact with each other is to run realistic simulations. They want researchers to drop AI agents into sandboxes and study what they do.
You can’t predict what’s going to happen by studying single agents, or even small groups of agents, in isolation. You can’t assume that AI agents underpinned by LLMs will always act rationally, says Fox. And the complexity comes from having huge numbers of interactions at once.
Google DeepMind is not the only top AI firm warning about the risks of the technology it is building. A couple of weeks ago, Anthropic published guidelines for deploying AI agents based on an approach to cybersecurity known as zero trust, which starts with the assumption that a computer system is vulnerable, an agent is an attacker, and a breach will happen.
Refael Angel, cofounder and CTO of Akeyless, a cybersecurity firm based in Tel Aviv, agrees that understanding the new risks introduced by agent-based systems is crucial.
Every approach to security in the past has assumed that the machine in question was software written by a human, doing fixed things on fixed paths, says Angel: “An agent breaks all of those assumptions. It reasons, it improvises, and it can be hijacked by a single sentence buried in a document it was asked to read.”
Angel welcomes this new funding. “No single lab should author the safety standards everyone else has to trust,” he says. But he cautions that safety researchers can overlook boring problems that are already here in favor of more exotic hypothetical ones.
And yet, Fox notes, risks that were hypothetical a few years ago are now very real: “The future’s come more quickly than perhaps expected.”
Background: Adolescent sleep duration can substantially impact mood, behavior, and academic attainment. While hundreds of sleep-related apps are available to download, none have been cocreated with adolescents from underserved populations in the United Kingdom. Objective: This study aimed to explore adolescents’ views, expectations, and experiences with a novel app to improve sleep, called Sleep Solved, to understand which features were perceived as positive and helpful, and to identify ways to further enhance its usefulness. Sleep Solved is part of a larger stepped behavior change study and was cocreated with adolescents from underserved groups to make the app accessible and engaging for this population. Methods: A total of 63 participants aged 16‐18 years from across the United Kingdom completed semistructured interviews after trying the app. Interviews were analyzed using inductive thematic analysis, as outlined by Braun and Clarke, with a particular focus on the views of individuals from underserved ethnic and socioeconomic groups. Results: Participants perceived Sleep Solved as a useful tool that provides helpful advice regarding changeable behaviors to improve sleep hygiene. Cocreated features of the app, such as the Sleep Stars gamified rewards system and the easy-read, science-based “sleep hacks,” were viewed positively by participants, who reported that they had a beneficial impact on their sleep and sleep schedule. Praise was given for the app’s ease of use and how the science of sleep was explained at an appropriate level, without being overwhelming. Compared to sleep advice on social media platforms, Sleep Solved was considered more reliable and trustworthy. Participants described better sleep hygiene, such as a regular sleep routine and a longer sleep duration, and increased feelings of improved mood and energy. Conclusions: This study found that a cocreated sleep app, designed with input from adolescents in underserved UK populations, was perceived as accessible, reliable, and effective in supporting positive sleep behavior change. Although sleep duration was not objectively tested, participants, particularly those from low socioeconomic status backgrounds and diverse ethnicities, reported improved sleep routines and mood, highlighting the potential of co-designed digital tools to engage and benefit adolescent users.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/21c2b21bed8d0a7a06bba5b35e5f34fb" />
GlaxoSmithKline (GSK) has agreed to acquire Nuvalent for $10.6 billion, the companies said, in a deal designed to strengthen the buyer’s cancer pipeline with Nuvalent’s precision oncology treatments—including three non-small cell lung cancer (NSCLC) therapies, of which two are under FDA review with decisions expected later this year.
Boston-based Nuvalent’s pipeline is headed by the ROS1 inhibitor zidesamtinib (NVL-520) and the ALK inhibitor eladalkib (NVL-655), which according to the company represent potential best-in-class, next-generation, highly selective treatments for NSCLC. Both are brain penetrant. The FDA has set target decision dates of September 18 for zidesamtinib and November 27 for neladalkib, both of which have been granted the agency’s Breakthrough Therapy and Orphan Drug designations.
Zidesamtinib is designed to treat NSCLC tumors driven by ROS1 that have developed resistance to currently available ROS1 inhibitors, including tumors with the prevalent G2032R “solvent front” resistance mutation. Zidesamtinib is selective in order to minimize CNS adverse events related to off-target inhibition of the tropomyosin receptor kinase (TRK) family, and potentially drive durable responses for patients with ROS1-mutant variants, Nuvalent says.
Eladalkib was created to address treating tumors driven by ALK that have developed resistance to first-, second-, and third-generation ALK inhibitors, including tumors with both single or compound treatment-emergent ALK mutations such as those involving the G1202R “solvent front” mutation. Eladalkib is also designed to avoiding TRK family inhibition and to treat brain metastases.
The third NSCLC asset of Nuvalent, NVL-330, is a HER2 inhibitor now under study in Phase I trials for HER2-altered NSCLC. In addition, Nuvalent’s pipeline includes an unspecified number of preclinical programs focused on “addressing the limitations of existing therapies for clinically proven kinase targets in oncology,” the company states on its website.
“Today’s acquisition is a multi-product deal, consistent with our approach to acquire assets that have clinically proven targets and meaningfully address an efficacy and/or tolerability gap,” GSK CEO Luke Miels said in a statement. “The two lead products are potential best-in-class assets that could launch this year if approved by the FDA and offer significant new treatment options to patients with two forms of non-small cell lung cancer.”
GSK investors were less enthusiastic as its shares on the London Stock Exchange on Monday dipped 0.5% to 1,903.50 pence. However, Nuvalent shares jumped 39% on Nasdaq to $123.25.
The $10.6 billion Nuvalent acquisition is the third largest merger-and-acquisition (M&A) deal announced this year, behind the €10.7 billion ($12.355 billion) cash buyout offer for Italian-based Recordati being pursued by CVC Capital Partners and Groupe Bruxelles Lambert, which aim to take the company private; and Sun Pharmaceutical Industries’ planned $11.75 billion purchase of Organon, the women’s health drug developer spun out of Merck & Co., in a deal expected to close in early 2027.
Immediate sales opportunities
The Nuvalent candidates, GSK added, present immediate new sales growth opportunities, improving profit contributions from 2027, and a platform in lung cancer for rapid expansion with GSK’s Ris-Rez, a B7-H3 targeted antibody-drug conjugate (ADC) now in Phase III clinical development.
In a presentation to investors after announcing a series of business updates on May 27, Nuvalent projected an ROS1+ NSCLC treatment could generate ~$1.4 billion to $2.1 billion in peak year sales, with about 40% of those sales (from ~$570 million to $855M million) expected to come from the U.S.—multiples above the ~$150 million in peak year sales attained in 2019 by Xalkori® (crizotinib), marketed by Pfizer and Merck KGaA.
An ALK+ NSCLC treatment would potentially be even more lucrative, Nuvalent said last month, with projected worldwide peak year sales ranging from ~$3.4 billion to $5 billion, of which the U.S. would account for 40% of sales, or between ~$1.35 billion and $2 billion—well above the $519 million in peak sales attained in 2023 by Alecensa® (alectinib), marketed by Genentech, a member of the Roche Group and created by Roche-owned Chugai Pharmaceutical.
“Since our founding, we have leveraged our deep expertise in chemistry and structure-based drug design to develop a portfolio of novel, potentially best-in-class kinase inhibitors. Our close collaboration with leading physician-scientists and patient advocates has driven remarkable enrolment, accelerating development and building confidence in the clinical profile of these drugs,” Nuvalent CEO James Porter, PhD, stated. “We’re excited that GSK has recognized the significant value these programs can offer patients and shares our vision for practice-changing innovation.”
Positive pivotal data
In announcing the acquisition, GSK cited positive pivotal data Nuvalent presented at the IASLC 2025 World Conference on Lung Cancer and the 2026 ASCO Annual Meeting. Data at both conferences showed potential best-in-class profiles for zidesamtinib and neladalkib, with both treatments designed to deliver longer effective treatment with better quality of life than current therapies, through high target-selectivity, durable treatment response, improved tolerability, enhanced blood-brain barrier penetration for tumor spread, and broader coverage of ALK and ROS1 mutations.
ROS1- and ALK-altered NSCLC primarily affect non-smoking adults aged 40-50, GSK and Nuvalent said—a patient population the companies described as uniquely defined and engaged.
GSK said it will commence a tender offer to acquire all of Nuvalent’s outstanding shares of Class A and Class B common stock at a purchase price of $124 per share in cash within 10 business days. The expected purchase price represents a 40% premium to the last closing price and a 26% premium to the 30 calendar day volume-weighted average price.
Net of cash acquired, GSK estimated its aggregate investment in Nuvalent to be $9.4 billion.
GSK said the acquisition will not change its 2026 full-year guidance range of 7-9% core operating profit and core EPS growth. The acquisition is expected to contribute to revenue growth from 2027, be incremental to GSK’s existing ambition for sales of >£40 billion (>$53.56 billion) by 2031, and strengthen the company’s core operating profit through the two-year period of loss of exclusivity for its aging blockbuster dolutegravir (2028-2030).
Dolutegravir is an HIV-1 integrase strand transfer inhibitor (INSTI) marketed as the monotherapy Tivicay® by Viiv Healthcare, in which GSK holds a 78.3% majority stake (and Shionogi, the remaining 21.7% after Pfizer cashed out its 11.7% stake, receiving $1.88 billion). Dolutegravir is also included in Viiv’s fixed-dose HIV combination therapies Dovato (dolutegravir and lamivudine) and Juluca (dolutegravir and rilpivirine).
Adding to core profit, EPS
GSK said it expected to add to its core operating profit in 2027 and core earnings per share (EPS) in 2029 by acquiring Nuvalent, even after accounting for cost-cutting synergies and “reprioritization,” which it defines as the shifting of personnel, capital, and other resources away from lower-yield, early-stage research or legacy programs toward higher-value clinical assets and corporate activities. Nuvalent reported 228 full-time employees, of which 144 are engaged in R&D, in its Form 10-K annual report for 2025, filed February 26.
Should the transaction close in Q3 2026 as expected, GSK said it expects low single-digit percentage dilution to core EPS this year through 2028.
The company said it will fund the Nuvalent acquisition primarily from new and existing debt facilities plus cash, with no impact expected to its credit rating. GSK ended Q1 with £3.442 billion ($4.608 billion) in cash and cash equivalents, up 1.3% from £3.397 billion ($4.548 billion) at the end of 2025.
The transaction is subject to customary closing conditions, including the tender of a majority of Nuvalent’s outstanding shares of Class A common stock in the tender offer and the expiration or termination of the applicable waiting period under the Hart-Scott-Rodino Act in the U.S. Soon after the closing of the tender offer, GSK expects to acquire any remaining shares of Nuvalent through a second-step merger under Delaware law at the same price per share.
GSK said it will account for the transaction as a business combination and assume Nuvalent’s existing revenue-sharing arrangements of low-single-digit royalties payable to Royalty Pharma and Deerfield. Royalty Pharma in December acquired for up to $315 million a pre-existing royalty interest in zidesamtinib and neladalkib from an undisclosed third party. Deerfield is Nuvalent’s largest shareholder.
“GSK’s proven track record, infrastructure, and expertise will support the successful commercialization of zidesamtinib and neladalkib, as well as accelerate advancement of our broader discovery pipeline,” Porter added.
