The human cell is wildly complex. Can AI decode it? | Silvana Konermann

0:00With no fees or minimums on checking accounts, it's no wonder the Capital One bank guy is so passionate about banking with Capital One. If he were here, he wouldn't just tell you about no fees or minimums. He'd also talk about how most Capital One cafes are open seven days a week to assist with your banking needs. Yep, even on weekends, it's pretty much all he talks about. In a good way. What's in your wallet? Terms apply. See CapitalOne.com slash bank. Capital One N.A. Member FDIC.

0:32Today's episode is sponsored by Kashi. Say hi to new Kashi Gut Health Granola, a straightforward, delicious way to fuel your gut made with prebiotics, fiber, and cultures. My favorite flavor, blueberry almond, has perfectly crispy clusters, delicious almonds, and bursts of whole blueberries. It's an easy fit for a daily routine, whether that's for breakfast or a snack. In a world of overwhelm, it's a straightforward, delicious choice. Try new Kashi Gut Health Granola. This episode is sponsored by We Share Health, an affordable alternative to traditional health insurance. While healthcare costs keep climbing every year, We Share is doing things differently.

1:13It's a non-profit built to put people first. No annual price hikes, no surprises. In fact, their membership prices haven't budged in five years. Members can pay up to 60 % less than they would for traditional health insurance, saving up to$5 ,000 per year. Plus members get real benefits, a nationwide PPO network with over 1 million physicians, unlimited telehealth, and access to more than 68 ,000 CVS Caremark pharmacies. WeShare goes beyond traditional health insurance, providing both extensive preventative and holistic care. Over 100 ,000 members have already made the switch. Call 1-888-WESHARE to get a free quote in minutes.

1:53That's 1-888-WESHARE. Or visit WeShareHealth.org to learn more. This is TED Health, a podcast from TED, and I'm Dr. Shoshana Ungerleiter. I'm a board-certified practicing internal medicine physician, and I'm your host for today's show. One of the things that I love most about medicine is that you never stop learning. I don't just mean keeping up with new studies or new treatments. I mean realizing over and over again how much there still is to discover about the human body. The longer I practice, the more I've noticed that certainty is actually pretty rare. We can replace joints, transplant organs, map the genome, and treat diseases that would have been unimaginable just a generation ago.

2:42Then there are these moments that take you right back to the drawing board. Someone develops an illness that doesn't follow the pattern you expected. A medication works beautifully for one person and doesn't help another. A disease can be described in extraordinary detail. And we're still left asking how it truly began. Medicine moves forward when people are willing to ask smarter questions and look for better answers. Technological innovation has always been key to that process. Sometimes that meant building a new microscope. Sometimes it meant inventing a new imaging technique or sequencing DNA for the first time.

3:20And each breakthrough gave us another way to look beneath the surface. Now it feels like we're standing at another one of those moments. We've talked before about artificial intelligence helping physicians interpret scans or recognize patterns in enormous amounts of medical data. Those applications are exciting, especially for people caring for patients every day. AI can transform the healthcare experience, but some experts believe its potential is even greater than we realize. Living systems are astonishingly complex. Every cell is responding to many signals at once, and diseases like Alzheimer's, heart disease, and most cancers rarely have a single cause.

4:06They're the product of countless small interactions unfolding over years. Researchers are asking whether AI can help untangle some of that complexity. Here's the idea. Instead of looking at each experiment on its own, these AI systems can help connect results across thousands of experiments, gradually building a much bigger picture of how cells behave and communicate with each other. We're still early here. There are plenty of questions left to answer. and no one knows exactly how this work will reshape medicine. But innovations like these expand the kinds of questions we're able to ask. And that's often how science moves forward.

4:48We don't suddenly know everything. We simply become capable of seeing something we couldn't before. That's why I'm excited to share this conversation between Silvana Konerman and Chris Anderson, the former head of TED. Dr. Konerman is a bioengineer and neuroscientist, an assistant professor of biochemistry at Stanford University. She's also the co-founder and executive director of ARC Institute, a nonprofit organization for AI and biology research. Their conversation explores where AI and biology meet and how that intersection could change the way we understand disease and discover new treatments.

5:26And now, Sylvana Konerman and Chris Anderson in conversation.

5:34Great to see you. Welcome to TED. Now, Silvana, you've been passionate about science for quite a long time. Tell me about this picture. Yeah, so this picture is when I was 15. So I was born in a small town in Switzerland. My parents weren't into science, but somehow I got really fascinated just with nature around me and also just how we work as humans in our biology. So I really wanted to find a way to be able to get into a lab to do some science. It was actually pretty tricky for me, but eventually I talked one of my science teachers into convincing one of his colleagues to let me go into the lab.

6:10And so this is me with that first science project where I went on to win the national competition and then also the European Union competition. And I think that's really where I got, I think, the confidence to continue with science since then. But there's a drawback to being a scientific prodigy, which is that you end up feeling like you might have a responsibility to do something with that. And I think you've had that your whole life and you've thought about, you know, what are the biggest problems you could work on? Yeah, I have been, you know, I guess doing science now for more than 20 years, but I did want to say this is actually my first real public appearance.

6:48So I am very much usually behind the scenes. But yeah, I think a problem that I've really been thinking a lot about, I would say since undergrad, I did my undergrad in Switzerland, and in biology and neuroscience, I learned more about Alzheimer's disease. And we're learning how there are these big changes in the brain that are happening. A lot of it is known about late stages of the disease, how severe it is. And then also the lecture, though, ended with basically, but we have no idea really how it's starting. We still don't have a therapy. And that was now, you know, a long time ago. That was, I guess, you know, 17 years ago or so.

7:34And that really stuck with me because it felt like, why didn't we understand how it's starting? Why didn't we have a therapy? Like, there are all these, like, very observable changes happening. And so that got me interested in disease biology, and specifically complex diseases, where Alzheimer's is a complex disease. What that means is it sounds, oh, it's just complicated, but that's not what it means. It means that there are multiple different risk factors, and basically every patient has a unique combination of risk factors for a disease. That's different from an infection where you have one cause.

8:09And several of these diseases are similarly fundamentally complex. That's right. So heart disease, many cancers, obviously not accidents, but stroke and Alzheimer's disease. These are all complex diseases. And so that's why it's been so resistant to dramatic advancements in medical science in recent years. Basically, yeah. Basically, all of these have a combination of genetic changes, environmental factors, and each patient is unique. They have a unique combination of risk factors. And so we've been really struggling, I guess, as a scientific community, understanding what do all these different patients have in common that we could target and then fix the disease.

8:51But you're seeing now an opportunity to have a different kind of assault on these diseases. What has changed? I think there are now three things really that have come together really just in the last one or two years that make it possible to understand such a complex problem. like Alzheimer's disease and other diseases like it. And that's at the high-level three areas. If you kind of summarize it really quickly, it's measuring, changing, and understanding. And so measuring, what that means for us, is really single-cell sequencing. So this is a technology that allows us to look at one cell at a time and take a snapshot of key dynamic process in the cell, which is the RNA expression of the cell.

9:33So it's basically RNA is like the language of the cell, And this takes a snapshot, one cell at a time, of what's going on inside it. And then the second step, which is changing, we need to have the ability to change something very precise. So changing one gene at a time, stopping it from making the RNA, or changing it to upregulate the RNA. This is the area that I've been working on now for 15 years, CRISPR technology. And as a field, we've made a lot of advancements. and now we can do this across all the genes in the genome. We can make these changes in a targeted way. And it's really only possible also very recently.

10:11And then finally, I mean, of course, AI is at the forefront of everything, especially today. But we've just seen over, I would say, really the last two years that it's getting real, it's really working, and AI can help us understand these kinds of processes. So if I understand right, just as AI has cracked understanding human language, you see a possibility that AI could be used to understand the language of our own cells, RNA. Yeah, exactly. That's basically the core principle. And for that, you need to be able to measure it and change it in this targeted way. But as an analogy, the field was downing this at the time.

10:57I mean, even six years ago, it wasn't clear that people were not sure that you could really scale these large language models just based on language and kind of predicting language to actually build kind of a conception of the world, essentially, and at least approximate intelligence clearly pretty well, right? that this is the key insight for the last six years, which is that a model can learn so much just from human language. And similarly, we can apply that concept to RNA, which is basically the language of the cell, especially the dynamic language of the cell, because it's changing all the time.

11:36It reflects what's happening to the cell, but also it reflects the cell's genetics. Is it approximately the same level of complexity as human language, or much more so? It's hard to say, but I will say one key difference for me, and I think this is why AI can be so powerful for biology, is that human language was generated by humans, right? So we understand it, right? We came up with it. Arnie language or the biological language has evolved. It was not generated by humans. So it's basically impenetrable for us, right? But AI doesn't care. To try and crack it, I think you have to take the same stance of just getting huge amounts of data.

12:19Talk about that process. Yeah, absolutely. So, I mean, really what we learn, again, for large language models is just they're very hungry. They're very data hungry. And really, we've, you know, been generating data for these language models for, you know, thousands of years, right? They're using all human language which has been generated by over generations and civilizations. In biology, we don't have anything similar to that, right? Especially when you're thinking about, OK, we need these precise measurements, kind of one cell at a time, and we also need to know what actually happened to that cell, because we're trying to build a predictive model, a dynamic model that can predict how a cell will change when something happens to it.

13:02And so we need to generate that data set, and that's kind of really core to being able to build any useful model here. So give a sense of how you actually do this. Essentially, this is really combining those first two elements I was talking about, which is making a targeted change. In this case, we're using CRISPR technology to turn a gene off or to turn it on. And we're doing that one gene at a time for one cell at a time. And then we're measuring the outcome using single-cell RNA sequencing. We're capturing what happened to the cell. So you do what you call a perturbation of the cell. and then you measure the output.

13:39How many experiments like that do you need to do? What's your plan? Yeah, so our plan is to do at least a billion of these experiments. It's a lot of experiments over the next four years. And you're not talking about, like in software, you're talking about a billion actual biological... Yeah, they're all physical experiments. I mean, I'm a biologist, an experimental biologist. We're working with a lot of experiments in the lab. And yeah, the way that we can do this is kind of using some tricks that makes this much more scalable, right? We're not actually like running a billion little, you know, individual kind of reactions.

14:16We're able to use kind of different barcoding technologies to kind of run these experiments in, you know, bigger pools and then back out what happened, what we did to the cells. Okay, so if things work out as you hope, and I guess you're already seeing evidence that it's working out, once you gather that data, you're able to get from the model something truly amazing. Talk about that. Yeah, so just to give a sense of why I feel that we can do the billion experiments, is we've done about 60 million experiments so far. You've done 60 million, right. Yeah, so we feel pretty good. We can keep going.

14:53But yeah, the whole point of this is that we want to learn not just, OK, if I have this cell and I make this change, what happens to the cell? Really, you know, my motivation for generating this model is ultimately for human health. And so for that, we can now have a disease state. And importantly, this can be, for example, you know, a certain cell, an Alzheimer's disease, let's say it's an immune cell in the brain, microglia. And we can measure, you know, what that looks like, not just for one patient, but across many patients. And this data is out there, so we don't even have to generate it.

15:28And so we can see, OK, all these diseased cells, and we can have all the healthy cells, but again, across people. And then we can ask the model, OK, the model knows how to change cells, right? So what intervention, what genetic change, what chemical change do I need to make to convert all the diseased cells across all the patients with the same disease back to the healthy cells? So that's an amazing sort of prediction that this model... Like, if you truly understand the language of DNA, the model can predict something that medicine has never known before. Because the answer to doing that might be quite a complex series of interventions are needed for that cell.

16:07It's not like you just give it an aspirin. It could be that it's a complex combination of things, or it's really just a question even of picking the correct one, right? There's 20 ,000 possibilities, could be up or down, there's 40 ,000 possibilities. And normally, the way this target identification in biomedicine works today is really this kind of guess and check approach. So you have a hypothesis, one gene, then you're spending a few years on checking whether that's the right one, right? So if you have 40 ,000 things to pick from, even if you just have to pick a single one, that takes forever, right?

16:41And that's why we haven't cured these diseases yet. Hmm. So what are you going to do with this model as you gradually refine it? I mean, And if I understand, so you're thinking of this as basically a virtual cell, is what you're creating. Or it's almost more than that. It's like a universal virtual cell that researchers can, whatever cell they're working on, they can use your model. Talk about what you're planning to do with it. The whole point of it is that it is a universal virtual cell, which means that it needs to learn how to generalize to a new kind of cell or a new state of a cell, a new disease, for example, without having seen data, training data for that, you know, new cell type.

17:23So that is a very challenging task, and that's why we're, you know, really thinking hard about how to do these experiments. But ultimately, the vision is that this is actually real. So we have already built our first model that came out eight months ago. It's not very good. So, I mean, to be clear, it is state of the art, right? It's the best model at the time that was published. but it has a really long way to go still to be at the accuracy that I think it needs to be to be really useful but an interface that uses that model that we have today and so what you can do is you can say okay I have this cell that I'm starting with and then I want to change this about the cell and then it spits out different basically changes that you can make to the cell that are most likely to shift it the way you want.

18:13So you're not holding on to this yourself or licensing it to companies. You're making this generally available. That's right, yeah. So we have a few ways that we really want people... Thank you. People to be able to interact with it and also follow along, right? So one is we're going to be releasing this tool later this year for people to try. We'll give caveats like this is not very accurate or this is going to be 20 % accurate. but also we're going to iterate over the next four years. We're also hosting a virtual cell challenge every year for the whole community. We had 1 ,000 teams participating in the first one, and that's really to move the whole field forward to get to where I think we need to be.

18:58So this is amazing. The amazing work in your institute is really going to catalyze research worldwide because you're making this tool available. Some people looking at that may go, well, wait a sec, isn't that a little bit dangerous? Some of the people playing with this may not have humanity's best interests at heart. What do you say to that? Yeah, I mean, that's definitely a question that I think, as you know, we got during the audacious process. I think the key thing to keep in mind that this is really just for human cells. We could, in theory, someone could build this kind of tool for a virus, and I would say, don't do that.

19:32That's a bad idea. Because, yes, then you can absolutely use it to create something that would be dangerous. But this really just allows you to shift human cells into a different state. And I think that would be pretty difficult to abuse. And in principle, if a nasty virus did come along and this model is working properly, that's a way of giving us one of the quickest. Yeah, I mean, it will tell you, for example, you know how the virus is targeting this gene in the cell. Right now we know, OK, we know what happens to the cell when that's getting targeted. So, yeah, it will absolutely help us to defend.

20:07So here's your team. Yeah. Tell us about them. Yeah, so ARC really, you know, was only started in 2021. It's kind of when we decided to launch. 2022 is really when we got up and running. So it's only been four years, but we've grown a lot. We're over 300 people now. And, you know, I think one thing we really wanted to be able to achieve with ARC is to bring people together from different disciplines and have AI and biology under one roof in one institute. And we started just around the right time where we could see what machine learning was going to mean for biology. Sylvana, I got so excited to see the audacious community get behind this and really help you expand this vision.

20:50It's hard to imagine a bolder effort at really tackling what humanity needs and in making us all feel better about AI. So someone here who's got, in their family, they've got Alzheimer's or they've got heart disease or whatever. What would you say to them? Yeah, I mean, I would say that I really think biomedicine is going to transform for these kinds of diseases, right? Not maybe in three months, right? So you have to be a little patient. but I think that within four years, five years, we will be able to have these models that are accurate enough to be useful. And then it's a totally different way of doing biology.

21:31It's not kind of one hypothesis at a time, right? A field like Alzheimer's can get really bogged down by just focusing on one dominant hypothesis that might be wrong. And with these models, you can actually take like a comprehensive data-driven look, you know, out of all the things that we could be targeting with a drug, what's going to happen with all of them and then which of them is going to be the most effective one. It's just a totally different way of tackling the problem that I think is so exciting. Silvana, thank you for your incredible vision and for sharing it with us here. Thank you.

22:03It's really fantastic.

22:09That was Silvana Konerman in conversation with Chris Anderson at the TED conference in 2026. And that's our show for today. Thank you so much for listening. TED Health is a podcast from TED, and I'd love to hear your thoughts about this episode. Send me a message on Instagram at ShoshanaMD. This episode was produced by me, Shoshana Ungerleiter, and Jess Shane. Edited by Alejandra Salazar and fact-checked by Vanessa Garcia-Woodworth. Special thanks to Maria Lagis, Farah Degrange, Daniela Balarezo, Constanza Gallardo, Tanzika Sangmarniwang, and Roxanne Heilash.

23:08Yep. Even on weekends, it's pretty much all he talks about in a good way. What's in your wallet? Terms apply. See CapitalOne.com slash bank. Capital One N.A. Member FDIC. So you're not worried about hosting the whole family for the holidays? Why would I be? Wayfair had everything I needed to get the house holiday ready. What if Mike brings his cheerful dogs? My tree has tons of five-star reviews telling me how sturdy it is. But what if everyone's plus one brings a plus one? We'll see how extendable this extendable table really is. And extra chairs are on the way. Okay, what if there's a blizzard and everyone has to sleep over?

23:45Ready to go with extra bedding and towels. From guest-ready essentials to holiday decor you can trust, Wayfair delivers. Guessing is for game shows, not your business. Workday is the enterprise AI platform for HR, finance, and IT. With AI grounded in your context to get work done and done right. It's a new work day.