This transcript has been edited for clarity
Eric J. Topol, MD: This is really a great occasion to have a chance to meet with Priscilla Chan and Mark Zuckerberg and discuss the remarkable progress and dedication at the Chan Zuckerberg Initiative (CZI). CZI has been around for well over 5 years now. And it also coincides with the 10-year anniversary of you two getting married. To get it started, how did you conceive the idea of CZI — the purpose of which is to cure, prevent, or manage all diseases by 2100?
Priscilla Chan, MD, Co-Founder and Co-CEO, CZI: We both knew that we always wanted to give back. And we knew that we were going to put a lot of time and effort into making sure that we are doing our part in building a better future.
And I want to be clear, we chose really big sectors that we’re working in. Science. Education. Cure, prevent, or manage all disease. Huge, audacious goals. The part that we wanted to get smart on is where our niche was going to be, what we could bring that was differentiated in the field to help everyone make progress. And so for the past 5 years, we’ve really experimented with different models, different ways of building things — inside CZI’s four walls, outside of CZI’s four walls, partnering with folks all across the field to try to figure out where our niche is and where we can do our best work.
Topol: The part that’s dedicated to life science and medicine is pretty extraordinary. It was back in 1994 when I came across The Economist saying that all serious diseases would be cured by 2050. So at least you weren’t as bold as that. You said end of the century.
Mark Zuckerberg, Co-Founder and Co-CEO, CZI: No, and we don’t necessarily think that they are all going to be cured. That’s why it’s cure, be able to prevent upfront, or just be able to manage them as ongoing chronic diseases. The goal is that if we can help accelerate the field of science, then we can bring in that time when we get to the state that we are able to manage all these things better and create a better world for our kids’ generation and generations after that.
Chan: There’s a rule of thumb, Mark, that you like to share: You overestimate what you can do in a year.
Zuckerberg: People always overestimate what they can get done in a year or two and often dramatically underestimate what they can get done if they work consistently on something over a longer period of time. So in tech, they say 10 years. I think science is a somewhat longer timeframe.
But there are a lot of really brilliant people working in the field. We view our goal as basically building tools, especially with my background as an engineer who built this company and built operations at scale, being able to deliver tools that all scientists can use to accelerate toward the state of being able to either cure, prevent, or manage all diseases. It seems like it’s worth going at. And I don’t know, maybe 2050 is too ambitious, but I’m still optimistic that by the end of this century, we’ll have made a lot of progress toward it.
Topol: I thought that picking another 80-some years to work with gave you a little extra leeway compared to that earlier projection. So actually, creating tools is a really important point, because many people think it’s all about breakthroughs, whereas you’ve taken a different track. A really nice example is Ed Boyden and optogenetics, where you can actually see human biology in action. Could you give some examples of other tools that you’ve been backing that you see as promising?
Zuckerberg: Maybe I’ll talk about one that I find interesting and you can, too.
Zuckerberg: One of the first projects that we took on was helping to work with the community to pull together a standardized format for these cell atlases. And this is not just one project. The Biohub has been working on a number of different atlases for different organisms.
But at CZI, we’ve helped to build this tool, cellxgene, that people can use to browse, really efficiently, all of the different data that people are putting in. There have been some interesting cases around this, where if you look at autoimmune diseases, people want to see how immune cells are responding to this. We’ve had cases where scientists have been able to look and see how sets of a million cells or more respond to different things. It’s been really cool and powerful to see, as an example of a tool to help the field make progress.
Topol: And to that point, the Cell Atlas, which has a number of imminent major publications — just yesterday, in Nature Aging, they reported a frailty-specific monocyte. They took cells from the fetal stage all the way to elderly. So the periodic table of cells that you’ve come up with is an extraordinary tool, which I know is going to enable a lot of things.
Zuckerberg: This is clearly an effort that is much bigger than what we’re doing. The whole field — there are thousands of people working on this across different institutions around the world. But I think part of the role that we can play is through funding, helping to standardize some of the data formats and then use engineering to build tools to make it more easily accessible to people, and open access.
Those are two of the contributions there that we’re proud of. But it’s really important to us to not shortchange all the work that everyone out in the field is doing. We feel like we’re tool builders that help to arm people. The Cell Atlas isn’t a thing that we are doing.
Chan: The one that I was going to bring up is similarly built off of an existing field of open-source software, and folks building stuff, tinkering in their labs. So it’s like, I need a tool, I can put something together, and it’s napari. Napari is a tool to help analyze data that come out of the amazing microscopes that exist in universities and laboratories. Right now, without efficient and modern tools to pair with this phenomenal hardware, it can take scientists a really long time to be able to gather insights from the data they collect. So we’re building napari with the open-source community to help everyone be able to gather the insights and analysis that they need from the data.
But the thing that’s really fun and cool is to see the community that’s coming around the napari hub. Everyone has slightly different needs, and so they’re building plug-ins that come together on this shared platform that can meet individual needs, and a community of people are sharing Here’s how I work with this, here’s how this works. Because what we’ve seen is that people doing this work — creative, curious scientists — aren’t just going to take the tool at face value. They want to tinker. They want to understand how it works. They want to partner with someone else to try a new way of looking at things. And I think that’s really exciting.
Topol: What’s interesting is this fusion of Mark’s background in engineering and computing with yours in medicine, as a pediatrician. You each gave some examples of tools from the opposite space, yours from AI and Mark’s from biology. It’s striking how the two of you came together with different backgrounds, and it represents the future of life science and medicine in so many respects.
One thing that was really interesting is the philosophy you brought to CZI and the Biohub. It’s different from, say, the National Institutes of Health (NIH), which is our major funding source. You’re really into open science and open everything, like your support of bioRxiv and medRxiv to get all these pandemic preprints and everything else out, and napari, but you are also betting on talent. What I am impressed about with the Biohub is how investigators are selected based on their promise — young investigators, by the way, not necessarily the old folks, who have multiple grants already and a proven track record.
And other things, like the fact that it’s international. Almost 40% are from outside the United States, 40% are computing scientists, and 40% are less than 6 years into their careers. These are not the characteristics of other life science support entities. Could you talk about your philosophy when you got started and what you’ve learned in these first 5 years of the CZI and the Biohub? Is that working? The things that you selected — your big bets, if you will — are they clicking?
Chan: The Biohub is really interesting because it’s part of that culture change that is happening in science that we’re excited to be able to support. The Biohub investigators are early in their careers, diverse, and excited to think about how science can be done differently.
But one thing that they have selected themselves is that they are collaborative. They want to learn from someone who has completely different training from them. They all have different backgrounds. Sometimes you have a microbiologist working with an engineer. How do their fields come together? We have physicists coming in. How do they actually take what they know from their different disciplines to drive breakthroughs?
What you see there is that as in medicine, you get super-specialized, and sometimes you lose sight of the bigger picture and the knowledge that’s gained in other fields. The investigators at the Biohub are really excited about the fact that people from different walks of science help solve important problems. The other thing that has to be true is a lack of ego. They’re results-driven. They want to get the science down. So that’s been really fun to be able to support.
We found that academia plays an incredibly important role in breakthroughs. Industry does a lot in making sure that products reach patients. But there’s a space in the middle. The Biohub has helped with big problems that will need coordinated, long-term attention but don’t have immediate commercial value. And that bigger hive, if you will, that we’ve been able to bring together at the Biohub has driven incredible progress. That’s why we put out a request for proposals for Biohub number two, and we got 58 proposals from across the country.
I read them all. They were great. There’s a space in the ecosystem for that type of coordinated effort that we’re excited to support.
Topol: You just touched on the next 10 years, and there are three big components of that. There’s so much going on here.
You mentioned opening up more Biohubs in the years ahead. The one led by Steve Quake and Joe DeRisi has been phenomenal here in the Bay Area, with UCSF and Stanford and UC Berkeley. It’s great that 58 have applied for the next one. There will eventually be a network effect of all of them. But then, also, I understand, you’ve extended the current Biohub to 2031. So that gives a little extra runway.
Zuckerberg: Let’s talk about the theme for the next 10 years. Priscilla was just talking about how for the first 5 years, we were trying to figure out the different approaches we could take to implement these different values of long-term focus, collaboration among different people coming together, and integrating engineering and science together to build tools. We tried a bunch of different things, and some worked better than others.
We’re ready to basically double down on the things that have worked. We’re really proud of the imaging program, and we think that we should grow that. The Biohub has worked really well. So we want to see if we can replicate that a few more times and build out a network of Biohubs.
After the first 5 years, when we basically were just trying to see what kind of tools and mechanisms would work, now we’re embarking on a decades-long journey. And in every long-term problem that I’ve worked on, at the beginning of your work, it makes sense to set aside some time to instrument the problem and try to measure everything, and make sure that you’re making progress. So that way, for the work that you do for the majority of the time after that, you have a good foundation of measurement.
Topol: Another big thing that you’ve done is the Chan Zuckerberg Institute for Advanced Biomedical Imaging, which I’d like to hear more about. Obviously, AI, it’s the sweet spot of deep networks and imaging. Maybe you can just give more color about this — you’ve been on it already, but you’re taking it to another level.
Zuckerberg: For the next 10 years, a lot of the focus for the different work that we are going to fund is going to be about basically measuring human biology in action. The imaging institute is one example of this. (The Imaging Institute will be a part of the broader Imaging Program at CZI).
One of the major scientific focuses for the imaging institute is to be able to map out the location of every protein in a cell. So it gives you a sense of the range across different spatial scales and time that we’re trying to enable scientists to tackle. This is a long-term project. We’re going to need to be working on it for 10 or 15 years.
The way that I think about this, as an engineer, is if you’re trying to write some code and get it to work, but you don’t have the ability to debug it and step through, line by line, and see how it’s going and how things are interacting, it just takes you a long time. You’re kind of guessing. So having that tool of being able to see an immune cell interact with something it’s trying to fight, or watch cells replicate live, or watch proteins interact inside cells live, could be really valuable tools for scientists to be able, for example, to test different hypotheses and have a deeper understanding of how this works.
The imaging institute is one version of this. It builds off of some of the programs that Priscilla was talking about before, the napari work and napari hub. We’re going to keep on doing that. We’re really proud of the progress that the whole open-source community has made around that.
But we also want to take an approach that’s a little more integrated, like what we’ve seen with the Biohub, toward bringing together physicists and engineers and scientists and machine learning folks, because at some level, being able to observe things and label them very clearly is somewhat of a computer science problem as well. So you bring all these things together. That’s not going to be something that you can do in a few years. It’s inherently a longer-term, 10- to 15-year problem. So we are working on that, and setting aside the budget and focus, and creating stability for some really great people to know that they can work on it over that period of time. That’s one example of the type of thing that we want to support over the next decade in the overall focus of measuring human biology.
Topol: It’s ideal, because imaging is such a perfect substrate for machine eyes, if you will. Just to convey the power of this: If you show a retinal picture to the leading retinal experts in the world and ask, is this from a man or a woman, their chance of getting it accurate is 50%, whereas a deep neural network gets it right 98%-99% of the time.
And there’s something about the way that the inputs from these neural networks for seeing cells — sub-organelles, the whole works — and being able to define proteins at the atomic level like never before, that is really extraordinary. It’s getting into the granular level of structural biology. It’s very exciting. And it makes sense, of course, that you’re going to work heavily in the imaging world.
Zuckerberg: We’ve met so many great people who could potentially lead this, and the team has been really focused on that. So we’re excited; we think we’ll have some pretty good news on that soon around who’s leading it and the team.
It looks like that this imaging center is going to be based in the Bay Area. That’s not 100% finalized at this point. But it looks strongly like it’s going in that direction.
Topol: Another area that you’ve also delved into is the Kempner Institute, really bio-inspired AI, where you’re going back and forth between natural human intelligence and artificial intelligence. Maybe you can talk a bit about that.
Zuckerberg: Not all the work that we’re doing is in the Bay Area. The Kempner Institute is based on the East Coast in Cambridge at Harvard. The idea there is to study the foundational basis of intelligence by studying, at the same time, machine learning and the advances that are made there, and biological intelligence, and the crossover between the two, and try to understand a little more of a theoretical foundation for this. Because in all the machine learning work that we do, it’s a highly empirical field.
The results are extraordinary. And it’s also one of the areas where I’ve found that the results routinely outpace and do better than I’d even expect as a technologist. So the next 10 years are going to really be exciting in terms of what’s possible in imaging and these other areas.
But when you get down to it, at a fundamental level, people don’t really understand how this works — which is kind of a funny place to be in, where you can tinker with it and get it to work better, and it’s fairly predictable in terms of how it works, but you want to understand the principles underneath it. At least I have this belief that if you do understand the principles, then you can make even faster progress and make sure that you’re pointing it in a direction that will generally be good for humanity and the world.
That’s the goal of this center. There’s a lot of great work that can be done in AI that’s focused on the computational and machine learning side, but there is a space in the field to study the crossover with neuroscience and biological intelligence. To some degree, the advances in AI not only will give us tools to study all the neuroscience and other things, but they will also help us understand the way our brains work, and hopefully vice versa, in order to do a better job of building this.
Chan: With our increasing understanding of genetics and how genetics impact human disease, we still make a huge leap when we go from genetic change to clinical phenotype. There’s a huge gap of what’s going on. And we make a lot of assumptions or correlations like, we think we see this, we think this happens.
But what we’re trying to do with our strategy in the next 10 years is actually get a sense of, for example, there’s a ton of proteins — molecular pathways that all fire when something changes. And to the human eye or to even the brightest of scientists, it’s just chaos. You can look at it, but it looks like chaos. But if you take computation and the advances that computation have promised for the field, you can bring order to it. You can understand the perturbation that happened that may or may not be associated with the phenotype — the actual molecular phenotype of a disease — which brings us so much closer to being able to adjust for it or address the change.
The computation is both necessary and interesting. Like, can you later study the model to understand human biology differently? What is the model that actually made sense of what appears to be pure chaos in a way that helps us understand more predictable patterns in how the human body works?
Topol: It’s an impressive way to move this field forward. And another thing that you’ve incorporated, besides this pan-disciplinary model, is actually getting patient input. So it isn’t just human biology out there in some siloed way; you’re actually getting patient engagement. I think that’s really laudable. We can never do enough of that, because we’re ultimately trying to cure and prevent these diseases, and if you don’t have the patients involved, it’s missing a major piece.
Chan: The patient work is incredibly important. If we go back to the culture of science, making sure that patients have a voice in helping drive and set the research agenda is really important. I went to one of our early neurodegeneration conferences, and I heard a patient with Parkinson’s say, Everyone thinks about the tremor. And yeah, the tremor is there. But you know what really bothers me? I’m so tired. I just don’t have the energy.
That’s a really important insight. Increasingly, having patients engaged in your research actually makes your research go faster, because especially in rare disease, unless you are engaging with patients more broadly, you only have the few patients coming in who live close to an academic center. An engaged patient helps you understand new paradigms of medicine. And that’s important.
Rare disease is not rare. There are lots of people with rare diseases. As we increasingly move down the path of precision medicine, diseases that we sort of lump together today — depression, high blood pressure — can be broken down into more precise diagnoses.
When you think about it, if you are able to study the whole human spectrum, a lot of interesting biological insights come out of it, like a very famous one involving PCSK9 and how we can better address high cholesterol in patients. That’s a real important scientific breakthrough. The patient community is so rich in so many ways that help accelerate our ability to do science.
Topol: Yes, in the example you just gave on the genome editing space, you could, with one shot, keep bad cholesterol low for a lifetime in someone who was destined to have major vascular disease. There’s no shortage of impressive things going on right now in life science and medicine. And you’re right in the midst of it and accelerating it all.
Another example is an AI tool that was Breakthrough of the Year for Science for predicting protein structures from their amino acids.
Chan: So impressive. Unbelievable.
Topol: So far, 300,000 proteins have been predicted, and there will be 100 million by the end of this year — half of the known proteins in the universe. This tool has breakthrough potential. And it seems that the template you’ve been working on here is to develop those kinds of tools in a diverse way, to advance the field and eventually get to this destination — which is, as you say, bold, but I think actually attainable. You’re making a big bet to get there. You’ve given out over a billion dollars already for the CZI work. And I think it’s starting to really show.
Another example, just recently, is multiple sclerosis, a terribly debilitating disease. And to think that this is, in many people, caused by the Epstein-Barr virus. If we could have made that discovery years ago?
Chan: I saw that paper. It’s incredible.
Topol: So many things are happening now. Even in the midst of the distractions of a pandemic or a war with Russia and Ukraine, exciting things are happening in life science and medicine, and it’s a lot of the work that you’re supporting and organizing. It’s an extraordinary time. I’m the old dog in medicine. I’ve never seen anything quite like this. So the fact that you’re helping us in the biomedical community, I can’t say enough how welcome that is.
Some thoughts about where this is going. One thing that would help people at the individual level is if they had a virtual health coach. Their phone would talk to them and tell them they’re at risk for a condition, and it prevents that from ever happening. It’s the fantasy in medicine that we’ve never actualized. This takes multimodal AI, and it’s going to be a long road. But do you think that someday we will have guidance as to how to prevent conditions, for those who are willing?
Zuckerberg: I’m optimistic about that. This is something that we talk about a lot. This was actually the first conversation that you and I had a number of years ago. It really inspired me and probably set us on some of the direction that we’re going in now. The ability to have something that is relatively noninvasive, or something that you could put inside yourself, that could measure your health and instrument things on an ongoing basis. You could wave a wand over and get a readout. There are a bunch of technical problems to solve to do something like that. There are a bunch of regulatory and other challenges that you would need to work through to make sure it’s safe.
But it’s an interesting area for research. There’s quite a spectrum of tools that need to get built. The imaging stuff that we’re doing is really quite far down the path of scientific research tools. There are other tools that will help to cure, prevent, and manage diseases that might be a little bit more clinical over time. Today we’re a little bit more focused on the research side, but some of these things have crossover value. Like with rare diseases, you have people who are particularly motivated to try new tools or approaches, and that in turn gives the researchers and the doctors more data on how to approach things.
But I’m pretty optimistic about what you’re talking about. I’m not sure exactly when that will come. The field is steadily solving the problems that would stand in the way of things like that being able to exist.
Chan: Our bodies are so dynamic that part of doing that is understanding, over the course of a day, the full spectrum of normal human physiology. So a lot of the work that we’re doing, even the single-cell work and imaging, is explaining what is even possible. When you look at such a microscopic level, when you look that deeply, what is the full range of normal? What are variants that are of no concern, and what are the variants of concern? I’ve had to train myself to take off my physician hat, because as a physician, I only want the data if I know what I’m going to do with them. As a researcher, it is very important to understand the full range of the human experience to be able to then build those tools.
One thing that I am acutely aware of in biomedical research is that a lot of folks aren’t represented in the full spectrum in research. Right now, if you look at who participates in research, it’s predominantly middle-aged Caucasian men. You’re leaving out a lot of folks. Being a woman is not a rare disease. We should be able to understand how a woman’s body changes and evolves over the course of her lifetime. People of color are dramatically underrepresented in our research base. Very close to my heart, we have no idea what’s going on with kids. They’re not just little adults.
Making sure that as we build this base — as you called it, a periodic table, a very important reference on which we want to be able to build the next generation of medicine — we need to make sure that people are fully represented and we can actually reference back to their biology, and not an abstraction of a relatively small subset of people.
Topol: I completely concur with the importance of the diversity, equity, and inclusion features that you’ve been advocating at CZI, the Biohub, and everything that you’ve been doing.
One thing that is really extraordinary about this time that we live in is that these advances are occurring. For example, the mRNA vaccines — it seems like they just took 10 months, but they took 30 years. Cancer immunotherapy — great for certain types of cancer — took 30 years. What is impressive here is the long view, that you’ve basically devoted these first 5 years to some building blocks. You’re starting to get some yardsticks, some measurements of how you can develop milestones for the future. But you’re not trying to say that there’s going to be breakthroughs in 10 months. Maybe you can speak to that, because you’re thinking about your grandchildren and where this is all going to land many decades from now, right?
Chan: I would also put AlphaFold into that category. Overnight successes take a lot of work. A lot goes into every single breakthrough that happens overnight.
Zuckerberg: Yes, totally. We’re lucky in that we’re relatively young and doing this work. So we can easily have an outlook, hoping that we’re going to get to do this work for many decades. That’s partially a feature of starting sooner, rather than waiting until our careers were further along, to start with philanthropy. That’s one piece.
The other is looking at the landscape of funding for science overall. You mentioned the NIH, and it’s great. It provides a very important role in the field, which is providing a base level of funding to a very wide range of people who are doing good work across the country.
But if you were looking at what the portfolio of scientific investments should be, you’d want some amount that’s just a base that’s spread pretty broadly. But you’d also want a small number of projects that are relatively concentrated, but bigger and larger investments. And right now, in the field, one of the things that I’m excited about is there are starting to be more people who are funding work like this. I think that that’s really exciting, too.
But at least when we were getting started, there was a little bit of a gap there. So we’ll keep on evolving our approach and seeing what it looks like — the overall portfolio of the country’s or the world’s scientific investments, and where it looks like there might be gaps.
We’re very grateful for the work that everyone else is doing in the field. And just because other people are taking different approaches doesn’t mean that we think that that’s wrong. We’re taking this approach because some of the other approaches that are in some ways even more important are already being handled. So it opens up space to try different things.
Topol: Right. Well, it’s hard to cover all that you’ve been doing for the last 5 years and the aspirations going forward. Do you think we’ve hit the high points?
Chan: I think so. Talk to our team. They have a lot of exciting news.
Topol: Well, I can tell you that the moment I saw you had recruited Cori Bargmann as your science director, I knew you were off to the races. And obviously, Steve Quake and I are good friends. And you’ve selected some amazing people to help propel this forward. We — the biomedical community — are indebted to you for your investment in us. We hope that we’re going to come through to achieve that goal. I don’t know what will happen.
Chan: Me, neither. That’s the good part, though.
Topol: If we actually prevent, cure, and manage all diseases by the end of the century, then what?
Zuckerberg: There’s always going to be a next thing. But this is a good one to work on.
Topol: Terrific. Thanks very much.
Chan: Thank you.
Zuckerberg: Thank you.