#93 – Daphne Koller: Biomedicine and Machine Learning

So one day is a very long time, and I don't like to make predictions of the type we will never be able to do X because I think that's, you know, that smacks of hubris. It assumes that never in the entire eternity of human existence will we be able to solve a problem. That being said, curing disease is very hard because often times by the time you discover the disease, a lot of damage has already been done. And so to assume that we would be able to cure disease at that stage assumes that we would come up with ways of basically regenerating entire parts of the human body in a way that actually returns it to its original state. And that's a very challenging problem. We have cured very few diseases. We've been able to provide treatment for an increasingly large number, but the number of things that you could actually define to be cures is actually not that large. So I think that's it. There's a lot of work that would need to happen before one could legitimately say that we have cured even a reasonable number, far less all diseases.

I think it depends on which disease. I mean, there are ones where I would say we're maybe not quite at 100 because biology is really complicated and there's always new things that we uncover that people didn't even realize existed. So but I would say there's diseases where we might be in the 70s or 80s. And then there's diseases in which I would say it's probably the majority where we're really close to zero.

I think Alzheimer's is probably closer to zero than to 80. There are hypotheses, but I don't think those hypotheses have as of yet been sufficiently validated that we believe them to be true. And there is an increasing number of people who believe that the traditional hypothesis might not really explain what's going on. I would also say that Alzheimer's and schizophrenia and even type two diabetes are not really one disease. There are almost certainly a heterogeneous collection of mechanisms that manifest in clinically similar ways. So in the same way that we now understand that breast cancer is really not one disease, it is a multitude of cellular mechanisms all of which ultimately translate to uncontrolled proliferation, but it's not one disease. The same is almost undoubtedly true for those other diseases as well. And it's that understanding that needs to precede any understanding of the specific mechanisms of any of those other diseases. Now, in schizophrenia, I would say we're almost certainly closer to zero than to anything else. Type 2 diabetes is a bit of a mix. There are clear mechanisms that are implicated that I think have been validated that have to do with insulin resistance and such, but there's almost certainly there as well many mechanisms that we have not yet understood.

Those mechanisms are certainly overlapping. There's a well-known phenomenon that says that for most diseases other than childhood diseases, the risk for contracting that disease increases exponentially year on year every year from the time you're about 40. So obviously there is a connection between those two things. That's not to say that they're identical. There's clearly aging that happens that is not really associated with any specific disease. And there's also diseases and mechanisms of disease that are not specifically related to aging. So, I think overlap is where we're at.

And both are quite sad. Well, I mean, there's processes that happen as cells age that I think are contributing to disease. Some of those have to do with DNA damage that accumulates as cells divide where the repair mechanisms don't fully correct for those. There are accumulations of proteins that are misfolded and potentially aggregate and those two contribute to disease and contribute to inflammation. There is a multitude of mechanisms that have been uncovered that are sort of wear and tear at the cellular level that contribute to disease processes, and I'm sure there is many that we don't yet understand.

Again, I think immortal is a very long time and I don't know that that would necessarily be something that I would want to aspire to, but I think all of us aspire to an increased health span, I would say, which is an increased amount of time where you're healthy and active and feel as you did when you were 20, and we're nowhere close to that. People deteriorate physically and mentally over time and that is a very sad phenomenon. So I think a wonderful aspiration would be if we could all live to, you know, the biblical 120 maybe in perfect healthy high quality of life. I think that would be an amazing goal for us to achieve as a society. Now is the right age 120 or 100 or 150, I think that's up for debate, but I think an increased health span is a really worthy goal. And anyway, in a grand time of the age of the universe, it's all pretty short.

Up until now, I don't think it's played very much of a significant role because largely the data sets that one really needed to enable powerful machine learning methods, those data sets haven't really existed. There's been dribbs and drabs and some interesting machine learning that has been applied. I would say machine learning/data science, but the last few years are starting to change that. So we now see an increase in some large data sets but equally importantly an increase in technologies that are able to produce data at scale. It's not typically the case that people have deliberately proactively used those tools for the purpose of generating data for machine learning. To the extent that those techniques have been used for data production, they've been used for data production to drive scientific discovery and the machine learning came as a sort of byproduct second stage of, oh, you know, now we have a data set let's do machine learning on that rather than a more simplistic data analysis method. But what we are doing at Insitro is actually flipping that around and saying here's this incredible repertoire of methods that bioengineers, cell biologists have come up with. Let's see if we can put them together in brand new ways with the goal of creating data sets that machine learning can really be applied on productively to create powerful predictive models that can help us address fundamental problems in human health. So really focus to get make data the primary focus and the primary goal and find use the mechanisms of biology and chemistry to create the kinds of data set that could allow machine learning to benefit the most.

So before we dive into the details of that let me take a step back and ask when and where was your interest in human health born? Are there moments, events, perhaps, if I may ask, tragedies in your own life that catalyze this passion or was it the broader desire to help humankind? So, I would say it's a bit of both. So on I mean my interest in human health actually dates back to the early 2000s when a lot of my peers in machine learning and I were using data sets that frankly were not very inspiring. Some of us old-timers still remember the quote unquote 20 news groups data set where this was literally a bunch of text from 20 news groups, a concept that doesn't really even exist anymore. And the question was, can you classify which news group a particular bag of words came from? And it wasn't very interesting. The data sets at the time on the biology side were much more interesting both from a technical and also from an aspirational perspective. They were still pretty small, but they were better than 20 news groups. And so I started out I think just by wanting to do something that was more I don't know society useful and technically interesting and then over time became more and more interested in the biology and the human health aspects for themselves and began to work even sometimes on papers that were just in biology without having a significant machine learning component. I think my interest in drug discovery is partly due to an incident I had with when my father sadly passed away about 12 years ago. He had an autoimmune disease that settled in his lungs. And the doctors basically said well there's only one thing that we could do which is give him prednisone. At some point I remember doctor even came and said hey let's do a lung biopsy to figure out which autoimmune disease he has and I said would that be helpful would that change treatment said no there's only prednisone that's the only thing we can give him and I had friends who were rheumatologists who said the FDA would never approve prednisone today because the ratio of side effects to benefit is probably not large enough. Today we're in a state where there's probably four or five maybe even more it depends for which autoimmune disease but there are multiple drugs that can help people with autoimmune diseases many of which didn't exist 12 years ago and I think we're at a golden time in some ways in drug discovery where there's the ability to create drugs that are much more safe and much more effective than we've ever been able to before lacking is enough understanding of biology and mechanism to know where to aim that engine and I think that's where machine learning can help.

Sure. So an animal model for disease is where you create effectively it's what it sounds like. It's often times a mouse where we have introduced some external perturbation that creates the disease and then we cure that disease and the hope is that by doing that we will cure a similar disease in the human. The problem is that oftentimes the way in which we generate the disease in the animal has nothing to do with how that disease actually comes about in a human. It's what you might think of as a copy of the phenotype, a copy of the clinical outcome. But the mechanisms are quite different and so curing the disease in the animal which in most cases doesn't happen naturally. The mice don't get Alzheimer's, they don't get diabetes, they don't get atherosclerosis, they don't get autism or schizophrenia. Those cures don't translate over to what happens in the human and that's where most drugs fails just because the findings that we had in the mouse don't translate to a human. The disease in the dish models is a fairly new approach. It's been enabled by technologies that have not existed for more than 5 to 10 years. So for instance, the ability for us to take a cell from any one of us, you or me, revert that say skin cell to what's called stem cell status, which is what's called a pluripotent cell that can then be differentiated into different types of cells. So from that pluripotent cell one can create a neuron or a cardiomyocyte or a hepatocyte that has your genetics but that right cell type and so if there's a genetic burden of disease that would manifest in that particular cell type you might be able to see it by looking at those cells and saying oh that's what potentially sick cells look like versus healthy cells and understand how and then explore what kind of interventions might revert the unhealthy looking cell to a healthy cell. Now, of course, curing cells is not the same as curing people. And so there's still potentially a translatability gap, but at least for diseases that are driven say by human genetics and where the human genetics is what drives the cellular phenotype, there is some reason to hope that if we revert those cells in which the disease begins and where the disease is driven by genetics and we can revert that cell back to a healthy state, maybe that will help also revert the more global clinical phenotypes. That's really what we're hoping to do.

Yes. It is honestly before that happened, I think people would have predicted that to be possible.

It is incredible and cool it was much more I think finicky and bespoke at the early stages when the discovery was first made but at this point it's become almost industrialized. There are what's called contract research organizations, vendors that will take a sample from a human and revert it back to stem cell status and it works a very good fraction of the time. Now, there are people who will ask I think good questions. Is this really truly a stem cell or does it remember certain aspects of what of changes that were made in the human beyond the genetics? It's passed as a skin cell or it's passed in terms of exposures to different environmental factors and so on. So I think the consensus right now is that these are not always perfect and there is little bits and pieces of memory sometimes but by and large these are actually pretty good.

So the reversal is not as of this point something that can be done at the scale of tens of thousands or hundreds of thousands. I think total number of stem cells or IPS cells that are what's called induced pluripotent stem cells in the world I think is somewhere between five and 10,000 last I looked. Now again that might not count things that exist in this or that academic center and they may add up to a bit more but that's about the range. So it's not something that you could this point generate IPS cells from a million people but maybe you don't need to because maybe that background is enough because it can also be now perturbed in different ways. And some people have done really interesting experiments in for instance taking cells from a healthy human and then introducing a mutation into it using some of the using one of the other miracle technologies that's emerged in the last decade which is CRISPR gene editing and introduced a mutation that is known to be pathogenic. And so you can now look at the healthy cells and the unhealthy cells, the one with the mutation, and do a one-on-one comparison where everything else is held constant. And so you could really start to understand specifically what the mutation does at the cellular level. So the IPS cells are a great starting point. And obviously more diversity is better because you also want to capture ethnic background and how that affects things. But maybe you don't need one from every single patient with every single type of disease because we have other tools at our disposal.

Well, first of all, there's the variability that's driven simply by the fact that genetically we're different. So, a stem cell that's derived from my genotype is going to be different from a stem cell that's derived from your genotype. There's also some differences that have more to do with for whatever reason. Some people's stem cells differentiate better than other people's stem cells. We don't entirely understand why. So, there's certainly some differences there as well. But the fundamental difference and the one that we really care about and is a positive is that the genetics are different and therefore recapitulate my disease burden versus your disease burden.

Well, a disease burden is just if you think I mean it's not a well-defined mathematical term although there are mathematical formulations of it. If you think about the fact that some of us are more likely to get a certain disease than others because we have more variations in our genome that are positive of the disease, maybe fewer that are protective of the disease. People have quantified that using what are called polygenic risk scores which look at all of the variations in an individual person's genome and add them all up in terms of how much risk they confer for a particular disease and then they've put people on a spectrum of their disease risk. And for certain diseases where we've been sufficiently powered to really understand the connection between the many many small variations that give rise to an increased disease risk. There is some pretty significant differences in terms of the risk between the people say at the highest decile of this polygenic risk score and the people at the lowest decile. Sometimes those differences are you know factor of 10 or 12 higher. So there's definitely a lot that our genetics contributes to disease risk even if it's not by any stretch the full explanation.

So just to get a sense I don't know if it's easy to do but what kind of data is useful in this disease in a dish model like what are what's what's the source source of raw data information and also from my outsider's perspective sort of biology and cells are squishy things and then how do you connect a computer to that which sensory mechanisms I guess. So that's another one of those revolutions that have happened the last 10 years in that our ability to measure cells very quantitatively has also dramatically increased. So back when I started doing biology in you know the late '90s early 2000s that was the initial era where we started to measure biology in really quantitative ways using things like microarrays where you would measure in a single experiment the activity level what's called expression level of every gene in the genome in that sample and that ability is what actually allowed us to even understand that there are molecular subtypes of diseases like cancer where up until that point it's like oh you have breast cancer. But then we when we looked at the molecular data it was clear that there's different subtypes of breast cancer that at the level of gene activity look completely different to each other. So that was the beginning of this process. Now, we have the ability to measure individual cells in terms of their gene activity using what's called single cell RNA sequencing, which basically sequences the RNA, which is that activity level of different genes for every gene in a genome. And you could do that at single cell level. So, that's an incredibly powerful way of measuring cells. I mean, you literally count the number of transcripts. So, it really turns that squishy thing into something that's digital. Another tremendous data source that's emerged in the last few years is microscopy and specifically even super resolution microscopy where you could use digital reconstruction to look at subcellular structures sometimes even things that are below the diffraction limit of light by doing sophisticated reconstruction and again that gives you tremendous amounts of information at the subcellular level. There's now more and more ways that amazing scientists out there are developing for getting new types of information from even single cells. And so that is a way of turning those squishy things into digital data into beautiful data sets.

So I think there's different ways in which this data could potentially be used. Some people use it for scientific discovery and say, 'Oh, look, we see this phenotype at the cellular level.' So let's try and work our way backwards and think which genes might be involved in pathways that give rise to that. So that's a very sort of analytical method to sort of work our way backwards using our understanding of known biology. Some people use it in a somewhat more forward that would if that was backward this would be forward which is to say okay if I can perturb this gene does it show a phenotype that is similar to what I see in disease patients and so maybe that gene is actually causal of the disease so that's a different way and then there's what we do which is basically to take that very large collection of data and use machine learning to uncover the patterns that emerge from it. So for instance, what are those subtypes that might be similar at the human clinical outcome but quite distinct when you look at the molecular data? And then if we can identify such a subtype, are there interventions that if I apply it to cells that come from this subtype of the disease and you apply that intervention, it could be a drug or it could be a CRISPR gene intervention, does it revert the disease state to something that looks more like normal, happy, healthy cells? And so hopefully if you see that, that gives you a certain hope that that intervention will also have a meaningful clinical benefit to people. And there's obviously a bunch of things that you would want to do after that to validate that, but it's a very different and much less hypothesis-driven way of uncovering new potential interventions and might give rise to things that are not the same things that everyone else is already looking at.

That's how I feel too. That's kind of exciting because I'm most of my day spent with data sets that I guess closer to the news groups. So this is a kind of it just feels good to talk about in fact I don't almost don't want to talk to you about machine learning. I want to talk about the fundamentals of the data set which is an exciting place to be.

What would these with these approaches? What do you hope what kind of diseases can be helped? We mentioned Alzheimer's schizophrenia, type two diabetes. Can you just describe the various kinds of diseases that this approach can help? Well, we don't know. And I try and be very cautious about making promises about something, oh, we will cure X. People make that promise and I think it's I tried to first deliver and then promise as opposed to the other way around. There are characteristics of a disease that make it more likely that this type of approach can potentially be helpful. So for instance, diseases that have a very strong genetic basis are ones that are more likely to manifest in a stem cell derived model. We would want the cellular models to be relatively reproducible and robust so that you could actually get enough of those cells and in a way that isn't very highly variable and noisy. You would want the disease to be relatively contained in one or a small number of cell types that you could actually create in an in vitro in a dish setting. Whereas if it's something that's really broad and systemic and involves multiple cells that are in very distal parts of your body, putting that all in a dish is really challenging. So we want to focus on the ones that are most likely to be successful today with the hope I think that really smart bioengineers out there are developing better and better systems all the time so that diseases that might not be tractable today might be tractable in 3 years. So for instance 5 years ago these stem cell derived models didn't really exist. People were doing most of the work in cancer cells and cancer cells are very very poor models of most human biology because they're a they were cancer to begin with and b as you passage them and they proliferate in a dish they become because of the genomic instability even less similar to human biology. Now we have these stem cell derived models. We have the capability to reasonably robustly not quite at the right scale yet but close to derive what's called

organoids which are these teeny little sort of multicellular organ sort of models of an organ system. So there are cerebral organoids and liver organoids and kidney organoids and

Is that not like the coolest thing?

Um, and then I think on the horizon we're starting to see things like connecting these organoids to each other so that you could actually start—and there's some really cool papers that start to do that—where you can actually start to say, okay, can we do multi-organ system stuff? Um, there's many challenges to that. It's not easy by any stretch, but it might—I'm sure people will figure it out and in 3 years or 5 years there will be disease models that we could make for things that we can't make today.

That's the hope.

Mhm. So to jump topics a little bit, can you maybe tell the origin story of the history, the origin story of MOOCs, of Coursera, and in general your teaching to huge audiences on a very sort of impactful topic of AI in general.

So I think the origin story of MOOCs emanates from a number of efforts that occurred at Stanford University around, um, you know, the late 2000s where different individuals within Stanford, myself included, were getting really excited about the opportunities of using online technologies as a way of achieving both improved quality of teaching and also improved scale. And so, um, Andrew for instance led the Stanford Engineering Everywhere which was sort of an attempt to take 10 Stanford courses and put them online, um, just as video lectures. I led an effort within Stanford to take some of the courses and really create a very different teaching model that broke those up into smaller units and had some of those embedded interactions and so on, which got a lot of support from university leaders because they felt like it was potentially a way of improving the quality of instruction at Stanford by moving to what's now called the flipped classroom model. Um, and so those efforts eventually sort of started to interplay with each other and created a tremendous sense of excitement and energy within the Stanford community about the potential of online teaching and led in the fall of 2011 to the launch of the first Stanford MOOCs. The um,

Open online courses. So they did not come up with the acronym. I'm not particularly fond of the acronym, but it is what it is.

Probably. So anyway, those courses launched in the fall of 2011 and within a matter of weeks, with no real publicity campaign, just a New York Times article that went viral, about 100,000 students or more in each of those courses. And I remember this conversation that Andrew and I had which is like, wow, this is just—there's this real need here. And I think we both felt like, sure, we were accomplished academics and we could go back to our labs, write more papers, but if we did that then this wouldn't happen. And it seemed too important not to happen. And so we spent a fair bit of time debating, do we want to do this as a Stanford effort, kind of building on what we'd started, do we want to do this as a for-profit company? Do we want to do this as a nonprofit? And we decided ultimately to do it as we did with Coursera. Um, and so we started really operating as a company at the beginning of 2012. Um,

And the rest is history.

sort of globalization of learning. I think there is a hunger for learning that, you know, globalization is part of it, but I think it's just a hunger for learning. The world has changed in the last 50 years. It used to be that you finished college, you got a job, and by and large the skills that you learned in college were pretty much what got you through the rest of your job history. And yeah, you learned some stuff, but it wasn't a dramatic change. Today, we're in a world where the skills that you need for a lot of jobs today didn't even exist when you went to college. And many of the jobs that existed when you went to college don't even exist today or are dying. So part of that is due to AI, but not only, and we need to find a way of giving people access to the skills that they need today. And I think that's really what's driving a lot of this hunger.

So I think if we even take a step back, all of this started in trying to think of new ways to teach or new ways to sort of organize the material and present the material in a way that would help the education process, the pedagogy.