Ground Truths

*/ ?>

Latest Episodes:

Katerina Akassoglou: Blood Clots, Brain Inflammation, and Covid Oct 19, 2024

Show notes

Superimposed on an impressive body of work on the blood-brain-barrier and immune system, Prof Akassoglou and her collaborators just published an elegant study in Nature that centered on the direct binding os the SARS-CoV-2 spike protein to fibrin with marked downstream pro-inflammatory effects. The findings and potential treatments have implications beyond Covid, Long Covid to other neurologic diseases.Full videos of all Ground Truths podcasts can be seen on YouTube here. The audios are also available on Apple and Spotify.Transcript with links to audio and to relevant papers, graphicsEric Topol (00:07):Well, hello this is Eric Topol with Ground Truths, and with me today is Katerina Akassoglou. She is at the Gladstone Institute and she is a remarkable neuroimmunologist who has been doing extraordinary work for three decades to unravel the interactions between the brain, blood vessels and the role of inflammation. So Katerina, there's a lot to discuss, so welcome.Katerina Akassoglou (00:40):Thank you. Thank you so much. It's a great pleasure to join.By Way of BackgroundEric Topol (00:43):It's really interesting going back in your career. First of all, we're thankful that you immigrated here from Greece, and you have become one of the leading scientists in this discipline of important discipline of neuroimmunology, which is not just about Covid that we're going to talk about, but Alzheimer's and neurodegenerative diseases. This is a really big hot area and you're definitely one of the leaders. And what I was impressed is that all these years that you've been working on the integrity of the blood-brain barrier, the importance of fibrinogen and fibrin, and then comes along the Covid story. So maybe what we can do is start with that, which is you've made your mark in understanding this whole interaction between what can get into the brain, through the blood-brain barrier and incite inflammation. So this has been something that you've really taken to the extreme knowledge base. So maybe we can start with your work there before we get into the important seminal Nature paper that you recently published.Katerina Akassoglou (01:57):Yes, of course. So since very early on, I was still a graduate student when we made the first discovery and at the time was like mid-90s, so it was really ahead of its time. That dysregulation of cytokine expression in the brain of mice was sufficient to induce the whole cascade of events, triggering neurodegeneration, demyelination in pathological alterations, very reminiscent of multiple sclerosis pathology. And it was really hard to publish that study at the time because it was not yet accepted that this regulation of the immune system modeling the brain can be linked to neurodegeneration. So that was 1995 when we made that discovery, and I became really interested, what are the pathogenic triggers that actually polarized the immune cells in the brain? So with this, of course, this transgenic animal was expressing TNF, it was an artificially made animal that we made, but naturally what were the triggers that would polarize the innate immune cells? So I looked really early on in this mice and what I found was that the very first event was leaks of blood-brain barrier. It was opening of the blood-brain barrier in this mouse before inflammation, before demyelination, before neuronal loss. And this is really what shaped the question that, is it possible that these blood leaks that happened very early in the pathology, could this be the instigators of pathogenic inflammation in the brain?Eric Topol (03:34):Yeah. So in a way, you got at this question because of the chicken-and-egg and what happens first, and you got to the temporal saying, which happened first as you said, the leak before you could see evidence of inflammation and being able to study this of course in the experimental model, which you couldn't really do in people. And what I love about the description of your career, which has been quite extraordinary contributions is connecting the dots between the blood, the inflammatory response and the brain. Perhaps no one has done that like you have. And before we get into the recent paper, a lot of people are not aware that a year ago, a group in the UK known as PHOSP-COVID, they published a really important paper in Nature Medicine of over 1,800 people who were hospitalized with Covid and they found that fibrinogen was the best marker for cognitive deficits at 6 and 12 months (Figure below)(04:40):So that's just one of many papers, but it's a particularly well done study that already before you got into this work that recently published had emphasized fibrinogen. And by the way, again, having spent a lot of years in clots in the arteries, for me, we have to just get it down to fibrinogen plus thrombin gets you to fibrin. Okay, so fibrin is a major player here when fibrinogen is cleaved. So here we have the basis that you established, which is the fibrinogen leakage into the brain, activating inflammation, activating microglia, which like the macrophages of the brain and inciting the whole process. And before we close, I want to not just talk about Covid, but Alzheimer's too. But now let's get into the study that you did, [Fibrin drives thromboinflammation and neuropathology in COVID-19] which is striking, I mean really striking. And can you kind of take us through, because you not only demonstrated the importance of fibrin in inciting neuroinflammation in this model, but also how you could reverse it or prevent it. So this, and you looked at it in many different ways, this was a systematic approach. Maybe you can take us through how you were able to make such compelling evidence.The Multimodal EvidenceKaterina Akassoglou (06:09):Yes, thank you. First of all, thank you for bringing up the human relevance because this was also our inspiration for the work that we did in the Covid study. So as you mentioned in Covid patients, fibrinogen unbiased mass spec analysis was identified as the predictive biomarker for cognitive impairment in Long Covid patients. And this was in addition to also neuropathology data about the abundance of fibrin deposition in the brain. And these were studies that were done by NIH that have found deposition of fibrin in the brain and the reports for the abnormal and puzzling coagulation in Covid that is not setting other infections and also in many cases not always relating with the severity of symptoms. So even mild cases of Covid also had increased coagulation. I was really intrigued by this human, all this evidence in human data, and I thought that maybe the way that we're thinking about this, that it's systemic inflammation that drives the clotting.(07:24):Maybe there's another aspect to this. Maybe there is a direct effect of the virus with the coagulation cascade, and in this way maybe this can be an instigator of inflammation. So this was the original idea to be able to reconcile this data from the clinic about why do we have this prevalence of coagulopathy in Covid. And of course, the second question is, could this also be a driver of the disease? And of course, we're in a unique position because we have been studying this pathway now for over 20 years to have all the toolbox, the genetic toolbox, the pharmacologic toolbox to be able to actually really address these questions with genetic loss of function studies, with a blood innate immunity multiomics pipeline that we have set up in the lab. And of course, with preclinical pharmacology in our ABSL3 facility. So we had the infrastructure in place and the source in place to actually really dissect this question with both genetic tools as well as also technology platforms.Eric Topol (08:29):And you had in vivo imaging, you're the director of in vivo imaging for Gladstone and UCSF. So you do have the tools to do this.Katerina Akassoglou (08:38):Yes. The imaging that you mentioned is really important because this is, we employed that very early in our studies over now 15 years ago. And the reason was sometimes from snapshots of histopathology, you cannot really understand the sequence of events. So by being able to image these processes, both neuronal activity, microglia activation, infiltration of peripheral cells in the brain, this is how we could see the steps that what happens early on and to be able to answer these chicken-and-egg questions that you mentioned. So these were very, they're very important experiments, especially at the beginning because they were hypothesis driving and we were able to ask the right questions to drive our research program.Eric Topol (09:26):Now was the binding of the spike protein to one key site in fibrinogen, was that known before? [See outstanding Figure below from Trends in Immunology]Katerina Akassoglou (09:36):No, this was not known. So there was evidence that there are abnormal clots in Covid, but it was not known whether the spike protein would directly bind to protein to the coagulation cascade. So one of the key discoveries in our study was to use peptide array mapping and be able to identify not only the binding, but exactly the domains on fibrin that spike binds too. And what we found was two key domains, one the inflammatory domain and the other the plasmin binding site, which is important for fibrin degradation. So this suggested a potential dual deleterious role for this interaction, both by maybe affecting inflammation, but also delaying fibrinolysis, which is the degradation of this toxic protein from the brain. And indeed, we found that this interaction was responsible for all these two aspects, including decreased degradation, more inflammation, but also at the same time increased, increased coagulation. So it was a really pathogenic interaction.Eric Topol (10:47):Yeah, actually it's pretty striking. You have these two sites, the plasmin cleavage site of fibrinogen, which as you say, we knew there was a problem with clots. We knew that, but we didn't know exactly the spike protein how exactly it was implicated, particularly with fibrinogen. And then this other site, the CD11b-C18, now that's fancy for surface receptors of macrophages. And basically, this is critical because it's this microglia activation in the brain, and I know you saw it in the lungs as well through this other site that spike protein activated. So you had a twofer here of things that you discovered that the SARS-CoV-2 spike protein was capable of doing. This was a really big revelation. And then you also looked at mice that were genetically manipulated. So maybe you can, because before we get to your antibody monoclonal, the ways that you proved this were, I mean, one thing after another is really systematic. So maybe you can teach us about that.Thanks for reading Ground Truths! This post is public so feel free to share it.Establishing CausalityKaterina Akassoglou (12:08):Yeah, sure. So the first was about chemistry experiment. So this of course, we had to get to the next step to see is there any causality for this pathway. So we employed genetic loss of function studies and we had knockout mice, either fibrinogen knockout mice, this mice have all blood proteins except fibrinogen, and they have a delay in coagulation so they don't clot properly. But we also had a mutant mouse, which is a fibrinogen NK mouse. And this was a mutation only within this inflammatory domain that you mentioned, inflammatory domain that binds to C11b-C18. Other names for this is of course complement receptor 3, Mac-1 (αMβ2). It's the same, many names for this receptor, that as you mentioned, is expressed not only in microglial in the brain, but also peripheral immune cells including macrophages as well as also neutrophils which are CD11b expressing.(13:12):So we now have genetic models to be able to look at both complete depletion of fibrinogen, but also a very specific mutation and very selective mutation that only blocks the inflammatory properties without affecting the properties of fibrin in hemostasis. And these mice were made many years ago by a very close collaborator, Jay Degen at the University of Cincinnati. So what we found is that when we block either the inflammatory domain or we completely deplete fibrinogen, there was this profound protection after infection in internasal infection with the virus in lung inflammation. And this was both suppression of oxidative stress and this pathogenic inflammation in the lung, but also decreasing fibrosis, which has been associated with also Long Covid. And the surprise came from the transcriptomic data. So when we did transcriptomic analysis in this mice in the lungs, we found perhaps the expected decrease in the immune signatures in macrophages. This was in line with our previous work in, as you mentioned, Alzheimer's models, multiple sclerosis models. But what also was really surprising is there was that genes that are associated with activation of NK cells were upregulated. And of course this was the first time we had infected these mice, previously we had not done an infection before. So I think that maybe because of this region we had not seen before in our data this immunomodulatory role of fibrin that not only surprises the macrophage response, but also increases these NK cells that are important for viral clearance.Eric Topol (15:00):So again, the finding another important unique finding is the natural killer (NK) cells and effect there from the activation of this, as you said, the inflammation site or the CD11b-C18 that we've been talking about. So now another layer of this, a dimension of your Nature paper was that you tested an antibody that you already had developed so-called 5B8. A monoclonal that specifically binds to the domain of the one we're talking about this inflammation domain of fibrinogen. So can you tell us about what that showed?Katerina Akassoglou (15:45):Yes, so we tested this antibody in different models of Covid, which were both models with neuroinvasion and models without neuroinvasion. So we used both transgenic mice for hACE2, the human ACE2 infected with Delta, but we also use mouse adapted viruses like Beta that is just in the wild type mice with no transgenic being involved that these are without neuroinvasion. And we wanted to see if the antibody had any potential protective effects. And what we found is that the antibody protected from inflammation in the lung. So the data looked so similar with a genetic mutation of this pathway, protection from inflammation, decreased fibrosis, increased viral clearance, so decreased spike and viral proteins in the lungs. But we also found a protection in the brain. So the brains of this mice, including both the models we used with neuroinvasion and without, they both have had microglia activation in the brain. And we also found neuronal loss in the Delta infected mice and the antibody protected from both neuroinflammation but also improved neuronal survival in the mice. Showing that there can be this despite regardless of which model we used, there was this protective effect suggesting that by blocking fibrin, either the periphery or in the brain, this could be protected for these models.Eric Topol (17:28):Yeah, so I mean this is fascinating because until now, until this report of yours and your colleagues at Gladstone, there was knowledge that there would be neuroinflammation from Covid, both in patients from various biomarkers and imaging as well as in experimental model. But what this did was take it to the fibrin story, and I guess that's one of the questions you nailed that how important fibrin is, but that doesn't necessarily rule out other triggers of neuroinflammation, right?Katerina Akassoglou (18:04):Oh, absolutely not. So I think that this is one of the mechanisms that can be very important, especially in…

Full show notes at the publisher

Patrick Hsu: A Trailblazer in Digital Biology Oct 13, 2024

Show notes

When I think of digital biology, I think of Patrick Hsu—he’s the prototype, a rarified talent in both life and computer science, who recently led the team that discovered bridge RNAs, what may be considered CRISPR 3.0 for genome editing, and is building new generative A.I. models for life science. You might call them LLLMs-large language of life models. He is Co-Founder and a Core Investigator of the Arc Institute and Assistant Professor of Bioengineering and Deb Faculty Fellow at the University of California, Berkeley.Above is a brief snippet of our conversation. Full videos of all Ground Truths podcasts can be seen on YouTube here. The audios are also available on Apple and Spotify.Here’s the transcript with links to the audio and external links to relevant papers and things we discussed.Eric Topol (00:06):Well hello, it's Eric Topol with Ground Truths and I'm really delighted to have with me today Patrick Hsu. Patrick is a co-founder and core investigator at the Arc Institute and he is also on the faculty at the University of California Berkeley. And he has been lighting things up in the world of genome editing and AI and we have a lot to talk about. So welcome, Patrick.Patrick Hsu (00:29):Thanks so much. I'm looking forward to it. Appreciate you having me on, Eric.The Arc InstituteEric Topol (00:33):Well, the first thing I'd like to get into, because you're into so many important things, but one that stands out of course is this Arc Institute with Patrick Collison who I guess if you can tell us a bit about how you two young guys got to meet and developed something that's really quite unique that I think brings together investigators at Stanford, UCSF, and Berkeley. Is that right? So maybe you can give us the skinny about you and Patrick and how all this got going.Patrick Hsu (01:05):Yeah, sure. That sounds great. So we started Arc with Patrick C and with Silvana Konermann, a longtime colleague and chemistry faculty at Stanford about three years ago now, though we've been physically operational just over two years and we're an independent research institute working at the interface of biomedical science and machine learning. And we have a few different aspects of our model, but our overall mission is to understand and treat complex human diseases. And we have three pillars to our model. We have this PI driven side of the house where we centrally fund our investigators so that they don't have to write grants and work on their very best ideas. We have a technical staff side of the house more like you'd see in a frontier AI lab or in biotech industry where we have professional teams of R&D scientists working cross-functionally on higher level organizational wide goals that we call our institute initiatives.(02:05):One focused on Alzheimer's disease experimentally and one that we call a virtual cell initiative to simulate human biology with AI foundation models. And our third pillar over time is to have things not just end up as academic papers, but really get things out into the real world as products or as medicines that can actually help patients on the translational side. And so, we thought that some really important scientific programs could be unlocked by enabling new organizational models and we are experimenting at the institutional scale with how we can better organize and incentivize and support scientists to reach these long-term capability breakthroughs.Patrick, Patrick and SilvanaEric Topol (02:52):So the two Patrick’s. How did you, one Patrick I guess is a multi-billionaire from Stripe and then there's you who I suspect maybe not quite as wealthy as the other Patrick, how did you guys come together to do this extraordinary thing?Patrick Hsu (03:08):Yeah, no, science is certainly expensive. I met Patrick originally through Silvana actually. They actually met, so funny trivia, all three Arc founders did high school science together. Patrick and Silvana originally met in the European version of the European Young Scientist competition in high school. And Silvana and I met during our PhDs in her case at MIT and I was at Harvard, but we met at the Broad Institute sort of also a collaborative Harvard, MIT and Harvard hospitals Institute based in Kendall Square. And so, we sort of in various pairwise combinations known each other for decades and worked together for decades and have all collectively been really excited about science and technology and its potential to accelerate societal progress. Yet we also felt in our own ways that despite a lot of the tremendous progress, the structures in which we do this work, fund it, incentivize it and roll it out into the real world, seems like it's really possible that we'll undershoot that potential. And if you take 15 years ago, we didn't have the modern transformer that launched the current AI revolution, CRISPR technology, single-cell, mRNA technology or broadly addressable LNPs. That’s a tremendous amount of technologies have developed in the next 15 years. We think there's a real unique opportunity for new institutes in the 2020s to take advantage of all of these breakthroughs and the new ones that are coming to continue to accelerate biological progress but do so in a way that's fast and flexible and really focused.Eric Topol (04:58):Yeah, I did want to talk with you a bit. First of all before I get to the next related topic, I get a kick out of you saying you've worked or known each other for decades because I think you're only in your early thirties. Is that right?Patrick Hsu (05:14):I was lucky to get an early start. I first started doing research at the local university when I was 14 actually, and I was homeschooled actually until college. And so, one of the funny things that you got to do when you're homeschooled is well, you could do whatever you want. And in my case that was work in the lab. And so, I actually worked basically full time as an intern volunteer, cut my teeth in single cell patch clamp, molecular biology, protein biochemistry, two photon and focal imaging and kind of spiraled from there. I loved the lab, I loved doing bench work. It was much more exciting to me than programming computers, which was what I was doing at the time. And I think these sort of two loves have kind of brought me and us to where we are today.Eric Topol (06:07):Before you got to Berkeley and Arc, I know you were at Broad Institute, but did you also pick up formal training in computer science and AI or is that something that was just part of the flow?Patrick Hsu (06:24):So I grew up coding. I used to work through problems sets before dinner growing up. And so, it's just something that you kind of learn natively just like learning French or Mandarin.New Models of Funding Life ScienceEric Topol (06:42):That's what I figured. Okay. Now this model of Arc Institute came along in a kind of similar timeframe as the Arena BioWorks in Boston, where some of the faculty left to go to Arena like my friend Stuart Schreiber and many others. And then of course Priscilla and Mark formed the Chan Zuckerberg Institute and its biohub and its support. So can you contrast for one, these three different models because they’re both very different than of course the traditional NIH pathway, how Arc is similar or different to the others, and obviously the goal here is accelerating things that are going to really make a difference.Patrick Hsu (07:26):Yeah, the first thing I would say is zooming out. There have been lots of efforts to experiment with how we do science, the practice of science itself. And in fact, I've recently been reading this book, the Demon Under the Microscope about the history of infectious disease, and it talks about how in the 1910s through the 1930s, these German industrial dye manufacturing companies like Bayer and BASF actually launched what became essentially an early model for industrial scale science, where they were trying to develop Prontosil, Salvarsan and some of these early anti-infectives that targeted streptococcus. And these were some of the major breakthroughs that led to huge medical advances on tackling infectious disease compared to the more academic university bound model. So these trends of industrial versus academic labs and different structures to optimize breakthroughs and applications has been a through current throughout international science for the last century.(08:38):And so, the way that we do research today, and that's some of our core tenets at Arc is basically it hasn't always been this way. It doesn't need to necessarily be this way. And so, I think organizational experiments should really matter. And so, there's CZI, Altos, Arena, Calico, a variety of other organizational experiments and similarly we had MRC and Bell Labs and Xerox PARCS, NIBRT, GNF, Google Research, and so on. And so, I think there are lots of different ways that you can organize folks. I think at a high level you can think about ways that you can play with for-profit versus nonprofit structures. Whether you want to be a completely independent organization or if you want to be partnered with universities. If you want to be doing application driven science or really blue sky curiosity driven work. And I think also thinking through internally the types of expertise that you bring together.(09:42):You can think of it like a cancer institute maybe as a very vertically integrated model. You have folks working on all kinds of different areas surrounding oncology or immunotherapy and you might call that the Tower of Babel model. The other way that folks have built institutes, you might call the lily pad model where you have coverage of as many areas of biomedical research as possible. Places like the Whitehead or Salk, it will be very broad. You'll have planned epigenetics, folks looking at RNA structural biology, people studying yeast cell cycle, folks doing in vivo melanoma models. It's very broad and I think what we try to do at Arc is think about a model that you might liken more to overlapping Viking shields where there's sort of five core areas that we're deeply investing in, in genetics and genomics, computation, neuroscience, immunology and chemical biology. Now we really think of these as five areas that are maybe the minimal critical mass that you would need to make a dent on something as complicated as complex human diseases. It's certainly not the only thing that you need, but we needed a critical mass of investigators working at least in these areas.Eric Topol (11:05):Well, yeah, and they really converge on where the hottest advances are being made these days. Now can you work at Arc Institute without being one of these three universities or is it really that you maintain your faculty and your part of this other entity?Patrick Hsu (11:24):So we have a few elements to even just the academic side of the house. We have our core investigators. I'm one of them, where we have dually appointed faculty who retain their latter rank or tenured appointment in their home department, but their labs are physically cited at the Arc headquarters where we built out a lab in Stanford Research Park in Palo Alto. And so, folks move their labs there. They continue to train graduate students based on whatever graduate programs they're formally affiliated with through their university affiliation. And so, we have nearly 40 PhD students across our labs that are training on site every day.(12:03):So in addition to our core investigators, we also have what we call our innovation investigators, which is more of a grant program to faculty at our partner universities. They receive unrestricted funding from us to seed a new project or accelerate an existing area in their group and their labs stay at their home campus and they just get that funding to augment their work. The third way is our technical staff model where folks basically just come work at Arc and many of them also are establishing their own research groups focusing on technology R&D areas. And so, we have five of those technology centers working in molecular engineering, multi-omics, complex cellular models, in vivo models, and in machine learning.Discovery of Bridge RNAsEric Topol (12:54):Yeah, that's a great structure. In fact, just a few months ago, Patrick Collison, the other Patrick came to Stanford HAI where I'm on the board and you've summarized it really well and it's very different than the other models and other entities, companies included that you mentioned. It's really very impressive. Now speaking of impressive on June 26, this past few months ago, which incidentally is coincident with the draft genome in the year 2000, the human sequence. You and your colleagues, perhaps the most impressive jump in terms of an Arc Institute contribution published two papers back-to-back in Nature about bridge RNA: [Bridge RNAs direct programmable recombination of target and donor DNA] and [Structural mechanism of bridge RNA-guided recombination.] And before I get you to describe this breakthrough in genome editing, some would call it genome editing 3.0 or CRISPR 3.0, whatever. But what we have today in the clinic with the approval of CRISPR 1.0 for sickle cell and thalassemia is actually quite crude. I think most people will know it's just a double stranded DNA cleavage with all sorts of issues about repair and it's not very precise. And so, CRISPR 2.0 is supposed to be represented by David Liu's contributions and his efforts at Broad like prime and base editing and then comes yours. So maybe you can tell us about it and how it is has to be viewed as quite an important advance.Patrick Hsu (14:39):The first thing I would say before CRISPR, is that we had RNA interference. And so, even before this modern genome editing revolution with programmable CRISPRs, we had this technology that had a lot of the core selling points as well. Any target will now become druggable to us. We simply need to reprogram a guide RNA and we can get genetic access to things that are intracellular. And I think both the discovery of RNA interference by Craig Mello and Andy Fire or the invention or discovery of programmable CRISPR technologies, both depend on the same fundamental biological mechanism. These non-coding guide RNAs that are essentially a short RNA search string that you can easily reprogram to retarget a desired enzyme function, and natively both RNAi and CRISPR are molecular scissors. Their RNA or DNA nucleases that can be reprogrammed to different regions of the genome or the transcriptome to make a cut.(15:48):And as bioengineers, we have come up with all kinds of creative ways to leverage the ability to make site specific cuts to do all kinds of incredible things including genome editing or beyond transcriptional up or down regulation, molecular imaging and so on and so forth. And so, the first thing that we started thinking about in our lab was, why would mother nature have stopped only RNAi and CRISPR? There probably are lots of other non-coding RNAs out there that might be able to be programmable and if they did exist, they probably also do more complicated and interesting things than just guide a molecular scissors. So that was sort of the first core kind of intuition that we had. The second intuition that we had on the technology side, I was just wearing my biology hat, I’ll put on my technology hat, is the thing that we call genome editing today hardly involves the genome.(16:50):It's really you're making a cut to change an individual base or an individual gene or locus. So really you're doing small scale single locus editing, so you might call it gene level or locus level cuts. And what you really want to be able to do is do things at the genome scale at 100 kb, a megabase at the chro…

Full show notes at the publisher

AI Snake Oil—A New Book by 2 Princeton University Computer Scientists Sep 24, 2024

Show notes

Arvind Narayanan and Sayash Kapoor are well regarded computer scientists at Princeton University and have just published a book with a provocative title, AI Snake Oil. Here I’ve interviewed Sayash and challenged him on this dismal title, for which he provides solid examples of predictive AI’s failures. Then we get into the promise of generative AI.Full videos of all Ground Truths podcasts can be seen on YouTube here. The audios are also available on Apple and Spotify.Transcript with links to audio and external links to key publications Eric Topol (00:06):Hello, it's Eric Topol with Ground Truths, and I'm delighted to welcome the co-author of a new book AI SNAKE OIL and it's Sayash Kapoor who has written this book with Arvind Narayanan of Princeton. And so welcome, Sayash. It's wonderful to have you on Ground Truths.Sayash Kapoor (00:28):Thank you so much. It's a pleasure to be here.Eric Topol (00:31):Well, congratulations on this book. What's interesting is how much you've achieved at such a young age. Here you are named in TIME100 AI’s inaugural edition as one of those eminent contributors to the field. And you're currently a PhD candidate at Princeton, is that right?Sayash Kapoor (00:54):That's correct, yes. I work at the Center for Information Technology Policy, which is a joint program between the computer science department and the school of public and international affairs.Eric Topol (01:05):So before you started working on your PhD in computer science, you already were doing this stuff, I guess, right?Sayash Kapoor (01:14):That's right. So before I started my PhD, I used to work at Facebook as a machine learning engineer.Eric Topol (01:20):Yeah, well you're taking it to a more formal level here. Before I get into the book itself, what was the background? I mean you did describe it in the book why you decided to write a book, especially one that was entitled AI Snake Oil: What Artificial Intelligence Can Do, What It Can't, and How to Tell the Difference.Background to Writing the BookSayash Kapoor (01:44):Yeah, absolutely. So I think for the longest time both Arvind and I had been sort of looking at how AI works and how it doesn't work, what are cases where people are somewhat fooled by the potential for this technology and fail to apply it in meaningful ways in their life. As an engineer at Facebook, I had seen how easy it is to slip up or make mistakes when deploying machine learning and AI tools in the real world. And had also seen that, especially when it comes to research, it's really easy to make mistakes even unknowingly that inflate the accuracy of a machine learning model. So as an example, one of the first research projects I did when I started my PhD was to look at the field of political science in the subfield of civil war prediction. This is a field which tries to predict where the next civil war will happen and in order to better be prepared for civil conflict.(02:39):And what we found was that there were a number of papers that claimed almost perfect accuracy at predicting when a civil war will take place. At first this seemed sort of astounding. If AI can really help us predict when a civil war will start like years in advance sometimes, it could be game changing, but when we dug in, it turned out that every single one of these claims where people claim that AI was better than two decades old logistic regression models, every single one of these claims was not reproducible. And so, that sort of set the alarm bells ringing for the both of us and we sort of dug in a little bit deeper and we found that this is pervasive. So this was a pervasive issue across fields that were quickly adopting AI and machine learning. We found, I think over 300 papers and the last time I compiled this list, I think it was over 600 papers that suffer from data leakage. That is when you can sort of train on the sets that you're evaluating your models on. It's sort of like teaching to the test. And so, machine learning model seems like it does much better when you evaluate it on your data compared to how it would really work out in the real world.Eric Topol (03:48):Right. You say in the book, “the goal of this book is to identify AI snake oil - and to distinguish it from AI that can work well if used in the right ways.” Now I have to tell you, it's kind of a downer book if you're an AI enthusiast because there's not a whole lot of positive here. We'll get to that in a minute. But you break down the types of AI, which I'm going to challenge a bit into three discrete areas, the predictive AI, which you take a really harsh stance on, say it will never work. Then there's generative AI, obviously the large language models that took the world by storm, although they were incubating for several years when ChatGPT came along and then content moderation AI. So maybe you could tell us about your breakdown to these three different domains of AI.Three Types of AI: Predictive, Generative, Content ModerationSayash Kapoor (04:49):Absolutely. I think one of our main messages across the book is that when we are talking about AI, often what we are really interested in are deeper questions about society. And so, our breakdown of predictive, generative, and content moderation AI sort of reflects how these tools are being used in the real world today. So for predictive AI, one of the motivations for including this in the book as a separate category was that we found that it often has nothing to do with modern machine learning methods. In some cases it can be as simple as decades old linear regression tools or logistic regression tools. And yet these tools are sold under the package of AI. Advances that are being made in generative AI are sold as if they apply to predictive AI as well. Perhaps as a result, what we are seeing is across dozens of different domains, including insurance, healthcare, education, criminal justice, you name it, companies have been selling predictive AI with the promise that we can use it to replace human decision making.(05:51):And I think that last part is where a lot of our issues really come down to because these tools are being sold as far more than they're actually capable of. These tools are being sold as if they can enable better decision making for criminal justice. And at the same time, when people have tried to interrogate these tools, what we found is these tools essentially often work no better than random, especially when it comes to some consequential decisions such as job automation. So basically deciding who gets to be called on the next level of like a job interview or who is rejected, right as soon as they submit the CV. And so, these are very, very consequential decisions and we felt like there is a lot of snake oil in part because people don't distinguish between applications that have worked really well or where we have seen tremendous advances such as generative AI and applications where essentially we've stalled for a number of decades and these tools don't really work as claimed by the developers.Eric Topol (06:55):I mean the way you partition that, the snake oil, which is a tough metaphor, and you even show the ad from 1905 of snake oil in the book. You're really getting at predictive AI and how it is using old tools and selling itself as some kind of breakthrough. Before I challenge that, are we going to be able to predict things? By the way, using generative AI, not as you described, but I would like to go through a few examples of how bad this has been and since a lot of our listeners and readers are in the medical world or biomedical world, I'll try to get to those. So one of the first ones you mentioned, which I completely agree, is how prediction of Covid from the chest x-ray and there were thousands of these studies that came throughout the pandemic. Maybe you could comment about that one.Some Flagrant ExamplesSayash Kapoor (08:04):Absolutely. Yeah, so this is one of my favorite examples as well. So essentially Michael Roberts and his team at the University of Cambridge a year or so after the pandemic looked back at what had happened. I think at the time there were around 500 studies that they included in the sample. And they looked back to see how many of these would be useful in a clinical setting beyond just the scope of writing a research paper. And they started out by using a simple checklist to see, okay, are these tools well validated? Does the training and the testing data, is it separate? And so on. So they ran through the simple checklist and that excluded all but 60 of these studies from consideration. So apart from 60 studies, none of these other studies even passed a very, very basic criteria for being included in the analysis. Now for these 60, it turns out that if you take a guess about how many were useful, I'm pretty confident most cases would be wrong.(09:03):There were exactly zero studies that were useful in a clinically relevant setting. And the reasons for this, I mean in some cases the reasons were as bizarre as training a machine learning model to predict Covid where all of the positive samples of people who had Covid were from adults. But all of the negative samples of people who didn't have Covid were from children. And so, essentially claiming that the resulting classifier can predict who has Covid is bizarre because all the classifier is doing is looking at the checks history and basically predicting which x-ray belongs to a child versus an adult. And so, this is the sort of error in some cases we saw duplicates in the training and test set. So you have the same person that is being used for training the model and that it is also used for evaluating the model. So simply memorizing a given sample of x-rays would be enough to achieve a very high performance. And so, for issues like these, I think all 60 of these studies prove to be not useful in a clinically relevant setting. And I think this is sort of the type of pattern that we've seen over and over again.Eric Topol (10:14):Yeah, and I agree with you on that point. I mean that was really a flagrant example and that would fulfill your title of your book, which as I said is a very tough title. But on page 29, and we'll have this in the post. You have a figure, the landscape of AI snake oil, hype, and harm. And the problem is there is nothing good in this landscape. So on the y-axis you have works, hype, snake oil going up on the y-axis. And on the x-axis, you have benign and harmful. So the only thing you have that works and that's benign is autocomplete. I wouldn't say that works. And then you have works facial recognition for surveillance is harmful. This is a pretty sobering view of AI. Obviously, there's many things that are working that aren't on this landscape. So I just would like to challenge, are you a bit skewed here and only fixating on bad things? Because this diagram is really rough. I mean, there's so much progress in AI and you have in here you mentioned the predicting civil wars, and obviously we have these cheating detection, criminal risk prediction. I mean a lot of problems, video interviews that are deep fakes, but you don't present any good things.Optimism on Generative AISayash Kapoor (11:51):So to be clear, I think both Arvind and are somewhat paradoxically optimistic about the future of generative AI. And so, the decision to focus on snake oil was a very intentional one from our end. So in particular, I think at various places in the book we outline why we're optimistic, what types of applications we think we're optimistic about as well. And the reason we don't focus on them is that it basically comes down to the fact that no one wants to read a book that has 300 pages about the virtues of spellcheck or AI for code generation or something like that. But I think I completely agree and acknowledge that there are lots of positive applications that didn't make the cut for the book as well. That was because we wanted people to come to this from a place of skepticism so that they're not fooled by the hype.(12:43):Because essentially we see even these positive uses of AI being lost out if people have unrealistic expectations from what an AI tool should do. And so, pointing out snake oil is almost a prerequisite for being able to use AI productively in your work environment. I can give a couple of examples of where or how we've sort of manifested this optimism. One is AI for coding. I think writing code is an application that I do, at least I use AI a lot. I think almost half of the code I write these days is generated, at least the first draft is generated using AI. And yet if I did not know how to program, it would be a completely different question, right? Because for me pointing out that, oh, this syntax looks incorrect or this is not handling the data in the correct way is as simple as looking at a piece of code because I've done this a few times. But if I weren't an expert on programming, it would be completely disastrous because even if the error rate is like 5%, I would have dozens of errors in my code if I'm using AI to generate it.(13:51):Another example of how we've been using it in our daily lives is Arvind has two little kids and he's built a number of applications for his kids using AI. So I think he's a big proponent of incorporating AI into children's lives as a force for good rather than having a completely hands-off approach. And I think both of these are just two examples, but I would say a large amount of our work these days occurs with the assistance of AI. So we are very much optimistic. And at the same time, I think one of the biggest hindrances to actually adopting AI in the real world is not understanding its limitations.Eric Topol (14:31):Right. Yeah, you say in the book quote, “the two of us are enthusiastic users of generative AI, both in our work and our personal lives.” It just doesn't come through as far as the examples. But before I leave the troubles of predictive AI, I liked to get into a few more examples because that's where your book shines in convincing that we got some trouble here and we need to be completely aware. So one of the most famous, well, there's a couple we're going to get into, but one I'd like to review with you, it's in the book, is the prediction of sepsis in the Epic model. So as you know very well, Epic is the most used IT and health systems electronic health records, and they launched never having published an algorithm that would tell when the patient was hospitalized if they actually had sepsis or risk of sepsis. Maybe you could take us through that, what you do in the book, and it truly was a fiasco.The Sepsis DebacleSayash Kapoor (15:43):Absolutely. So I think back in 2016/2017, Epic came up with a system that would help healthcare providers predict which patients are most at risk of sepsis. And I think, again, this is a very important problem. I think sepsis is one of the leading causes of death worldwide and even in the US. And so, if we could fix that, I think it would be a game changer. The problem was that there were no external validations of this algorithm for the next four years. So for four years, between 2017 to 2021, the algorithm wasn't used by hundreds of hospitals in the US. And in 2021, a team from University of Michigan did this study in their own hospital to see what the efficacy of the sepsis prediction model is. They found out that Epic had claimed an AUC of between 0.76 and 0.83, and the actual AUC was closer to 0.6, and AUC of 0.5 is making guesses at random.(16:42):So this was much, much worse than the company's claims. And I think even after that, it still took a year for sepsis to roll back this…

Full show notes at the publisher

Francis Collins: On Truth, Science, Faith and Trust Sep 11, 2024

Show notes

Francis Collins is a veritable national treasure. He directed the National Institutes of Health from 2009 to 2021. Prior to that he led the National Human Genetics Research Institute (NHGRI) from 1997-2009, during which the human genome was first sequenced. As a physician-scientist, he has made multiple seminal discoveries on the genetic underpinnings of cystic fibrosis, Huntington’s disease, neurofibromatosis, progeria, and others. This brief summary is barely scratching the surface oh his vast contributions to life science and medicine.A video clip from our conversation on hepatitis C. Full videos of all Ground Truths podcasts can be seen on YouTube here. The audios are also available on Apple and Spotify.Transcript with external inks and links to audioEric Topol (00:06):Well, I am really delighted to be able to have our conversation with Francis Collins. This is Eric Topol with Ground Truths and I had the chance to first meet Francis when he was on the faculty at the University of Michigan when I was a junior faculty. And he gave, still today, years later, we're talking about 40 years later, the most dazzling Grand Rounds during his discovery of cystic fibrosis. And Francis, welcome, you inspired me and so many others throughout your career.Francis Collins (00:40):Well, Eric, thank you and you've inspired me and a lot of other people as well, so it's nice to have this conversation with you in the Ground Truths format.Eric Topol (00:49):Well, thank you. We're at the occasion of an extraordinary book you put together. It's the fifth book, but it stands out quite different from the prior books as far as I can tell. It's called The Road to Wisdom: On Truth, Science, Faith and Trust, these four essential goods that build upon each other. And it's quite a book, Francis, I have to say, because you have these deep insights about these four critical domains and so we'll get into them. But I guess the first thing I thought I'd do is just say, how at some point along the way you said, “the goal of this book is to turn the focus away from hyperpartisan politics and bring it back to the most important sources of wisdom: truth, science, faith and trust, resting upon a foundation of humility, knowledge, morality, and good judgment.” So there's a lot there. Maybe you want to start off with what was in the background when you were putting this together? What were you really aiming at getting across?Reflections on CovidFrancis Collins (02:06):I'm glad to, and it's really a pleasure to have a chance to chat with you about this. I guess before Covid came along, I was probably a bit of a naive person when it came to how we make decisions. Yeah, I knew there were kind of wacky things that had gone out there from time to time, but I had a sort of Cartesian attitude that we were mostly rational actors and when presented with evidence that's been well defended and validated that most people will say, okay, I know what to do. Things really ran off the rails in the course of Covid. It was this remarkable paradox where, I don't know what you would say, but I would say the development of the vaccines that were safe and highly effective in 11 months using the mRNA platform was one of the most stunning achievements of science in all of history up until now.Francis Collins (03:02):And yet 50 million Americans decided they didn't want any part of it because of information that came to them that suggested this was not safe or there was conspiracies behind it, or maybe the syringes had chips that Bill Gates had put in there or all manner of other things that were being claimed. And good honorable people were distracted by that, lost their trust in other institutions like the CDC, maybe like the government in general like me, because I was out there a lot trying to explain what we knew and what we didn't know about Covid. And as a consequence of that, according to Kaiser Family Foundation, more than 230,000 people died between June of 2021 and April of 2022 because of a decision to reject the opportunity for vaccines that were at that time free and widely available. That is just an incredibly terribly tragic thing to say.Francis Collins (04:03):More than four times the number of people who died, Americans who died in the Vietnam War are in graveyards unnecessarily because we lost our anchor to truth, or at least the ability to discern it or we couldn't figure out who to trust while we decided science was maybe not that reliable. And people of faith for reasons that are equally tragic were among those most vulnerable to the misinformation and the least likely therefore, to take advantage of some of these lifesaving opportunities. It just completely stunned me, Eric, that this kind of thing could happen and that what should have been a shared sense of working against the real enemy, which was the SARS-CoV-2 virus became instead a polarized, divisive, vitriolic separation of people into separate camps that were many times driven more by politics than by any other real evidence. It made me begin to despair for where we're headed as a country if we can't figure out how to turn this around.Francis Collins (05:11):And I hadn't really considered it until Covid how serious this was and then I couldn't look away. And so, I felt if I have a little bit of credibility after having stepped down after 12 years as the NIH Director and maybe a chance to influence a few people. I just have to try to do something to point out the dangers here and then to offer some suggestions about what individuals can do to try to get us back on track. And that's what this book is all about. And yeah, it's called The Road to Wisdom because that's really how I want to think of all this in terms of truth and science and faith and trust. They all kind of give you the opportunities to acquire wisdom. Wisdom is of course knowledge, but it's not just knowledge, it's also understanding it has a moral character to it. It involves sophisticated judgment about difficult situations where there isn't an obvious answer. We need a lot more of that, it seems we’re at short supply.Deconvoluting TruthEric Topol (06:13):Well, what I really loved about the book among many things was how you broke things down in just a remarkably thoughtful way. So truth, you have this great diagram like a target with the four different components.in the middle, necessary truth. And then as you go further out, firmly established facts, then uncertainty and then opinion, and truth is not a dichotomous by any means. And you really got that down and you explained each of these different facets of truth with great examples. And so, this among many other things that you broke down, it wasn't just something that you read somewhere, you really had to think this through and perhaps this experience that we all went through, but especially you. But because you bring so much of the book back to the pandemic at times with each of the four domains, so that and the spider web. The spider web of where your core beliefsare and then the ones further out on the web and you might be able to work on somebody out further periphery, but it's pretty hard if you're going to get to them in the middle where their main thing is science is untrustworthy or something like that.Eric Topol (07:36):So how did you synthesize these because the graphics are quite extraordinary?Francis Collins (07:44):Well, I will say the artist for the graphics is a remarkable graphic design student at the University of Michigan who happens to be my granddaughter. So it was nice having that ability to have my scratches turned into something actually looks like artwork. The concepts I got to say, Eric, I was feeling pretty unsure of myself. I never took a course in philosophy. I know there are people who've spent their entire careers going all the way back to Socrates and on up until now about what does truth mean and here's this scientist guy who's trying to say, well, let me tell you what I think about it. I'm glad to hear that you found these circles useful. They have been very useful for me and I hadn't thought about it much until I tried to put it in some sort of framework and a lot of the problems we have right now where somebody says, well, that might be true for you, but it's not true for me, that's fine if you're talking about an opinion, like whether that movie was really good or not.Francis Collins (08:43):But it's not fine if it's about an established fact, like the fact that climate change is real and that human activity is the main contributor to the fact that we've warmed up dramatically since 1950. I'm sorry, that's just true. It doesn't care how you feel about it, it's just true. So that zone of established facts is where I think we have to re-anchor ourselves again when something's in that place. I'm sorry, you can't just decide you don't like it, but in our current climate and maybe postmodernism has crept in all kinds of ways we're not aware of, the idea that there is such a thing as objective truth even seems to be questioned in some people's minds. And that is the path towards a terrible future if we can't actually decide that we have, as Jonathan Rauch calls it, a constitution of knowledge that we can depend on, then where are we?Eric Topol (09:37):Well, and I never heard of the term old facts until the pandemic began and you really dissect that issue and like you, I never had anticipated there would be, I knew there was an anti-science, anti-vaccine sector out there, but the fact that it would become so strong, organized, supported, funded, and vociferous, it's just looking back just amazing. I do agree with the statement you made earlier as we were talking and in the book, “the development of mRNA vaccines for Covid in record time as one of the greatest medical achievements in human history.” And you mentioned besides the Kaiser Family Foundation, but the Commonwealth Fund, a bipartisan entity saved three million lives in the US, eighteen million hospitalizations. I mean it's pretty extraordinary. So besides Covid, which we may come back to, but you bring in everything, you bring in AI. So for example, you quoted the fellow from Google who lost his job and you have a whole conversation with Blake Lemoine and maybe you can give us obviously, where is AI in the truth and science world? Where do you stand there and what were you thinking when you included his very interesting vignette?Perspective on A.I.Francis Collins (11:17):Well, I guess I was trying to talk about where are we actually at the point of AGI (artificial general intelligence) having been achieved? That is the big question. And here's Blake Lemoine who claimed based on this conversation that I quote in the book between him and the Google AI apparatus called LaMDA. Some pretty interesting comments where LaMDA is talking about having a soul and what its soul looks like and it's a portal to all sorts of other dimensions, and I can sort of see why Blake might've been taken in, but I can also see why a lot of people said, oh, come on, this is of course what an AI operation would say just by scanning the internet and picking out what it should say if it's being asked about a soul. So I was just being a little provocative there. My view of AI, Eric, is that it's applications to science and medicine are phenomenal and we should embrace them and figure out ways to speed them up in every way we can.Francis Collins (12:17):I mean here at NIH, we have the BRAIN Initiative that's trying to figure out how your brain works with those 86 billion neurons and all their connections. We're never going to sort that out without having AI tools to help us. It's just too complicated of a problem. And look what AI is doing and things like imaging radiologists are going to be going out of business and the pathologists may not be too far behind because when it comes to image analysis, AI is really good at that, and we should celebrate that. It's going to improve the speed and accuracy of all kinds of medical applications. I think what we have to worry about, and I'm not unique in saying this, is that AI when applied to a lot of things kind of depends on what's known and goes and scrapes through the internet to pull that out. And there's a lot of stuff on the internet that's wrong and a lot of it that's biased and certainly when it comes to things like healthcare, the bias in our healthcare system, health disparities, inadequacies, racial inequities are all in there too, and if we're going to count on AI to fix the system, it's building on a cracked foundation.Francis Collins (13:18):So we have to watch out for that kind of outcome. But for the most part, generative AI it’s taking really exciting difficult problems and turning them into solutions, I'm all for it, but let's just be very careful here as we watch how it might be incorporating information that's wrong and we won't realize it and we'll start depending on it more than we should.Breathtaking AdvancesEric Topol (13:42):Yeah, no, that's great. And you have some commentary on all the major fronts that we're seeing these days. Another one that is a particularly apropos is way back when you were at Michigan and the years before that when you were warming up to make some seminal gene discoveries and cystic fibrosis being perhaps the first major one. You circle back in the book to CRISPR genome editing and how the success story to talk about some extraordinary science to be able to have a remedy, a cure potentially for cystic fibrosis. So maybe you could just summarize that. I mean that's in your career to see that has to be quite remarkable.Francis Collins (14:32):It is breathtaking, Eric. I mean I sort of like to think of three major developments just in the last less than 20 years that I never dreamed would happen in my lifetime. One was the ability to make stem cells from people who are walking around from a skin biopsy or a blood sample that are pluripotent. My whole lab studies diabetes, our main approach is to take induced pluripotent stem cells from people whose phenotypes we know really well and differentiate them into beta cells that make insulin and see how we can figure out how the genetics and other aspects of this determine whether something is going to work properly or not. I mean that's just astounding. The second thing is the ability to do single cell biology.Francis Collins (15:16):Which really 15 years ago you just had to have a bunch of cells and studying diabetes, we would take a whole eyelid and grind it up and try to infer what was there, ridiculous. Now we can look at each cell, we even can look at each cell in terms of what's its neighbor, does the beta cell next to an alpha cell behave the same way as a beta cell next to a duct? We can answer those questions, and of course the third thing is CRISPR and gene editing and of course the first version of CRISPR, which is the knockout of a gene was exciting enough, but the ability to go in and edit without doing a double stranded break and actually do a search and replace operation is what I'm truly excited about when it comes to rare genetic diseases including one that we work on progeria, which is this dramatic form of premature aging that is caused almost invariably by a C to T mutation in exon 11 of the LMNA gene and for which we have a viable strategy towards a human clinical trial of in vivo gene editing for kids with this disease in the next two years.Eric Topol (16:24):Yeah, it's just the fact that we were looking at potential cures for hundreds and pote…

Full show notes at the publisher

Joseph Allen: The Pivotal Importance of Air Quality, Ventilation and Exposures (Such as "Forever Chemicals") For Our Health Sep 05, 2024

Show notes

Professor Joseph Allen directs the Healthy Buildings Program at Harvard Chan School of Public Health. His expertise extends far beyond what makes buildings healthy. He has been a leading voice and advocate during the Covid pandemic for air quality and ventilation. He coined the term “Forever Chemicals” and has written extensively on this vital topic, no less other important exposures, which we covered In our wide-ranging conversation. You will see how remarkably articulate and passionate Prof Allen is about these issues, along with his optimism for solutions.A video snippet of our conversation: buildings as the 1st line of defense vs respiratory pathogens. Full videos of all Ground Truths podcasts can be seen on YouTube here. The audios are also available on Apple and Spotify.Transcript with External Links and Links to AudioEric Topol (00:00:06):Well, hello. It's Eric Topol from Ground Truths and I am just delighted to have with me, Joseph Allen from the Harvard School of Public Health, where he directs the Healthy Buildings Program that he founded and does a whole lot more that we're going to get into. So welcome, Joe.Joseph Allen (00:00:24):Thanks. It's great to be here. I appreciate the invitation.Joe Allen’s Background As A DetectiveEric Topol (00:00:28):Well, you have been, as I've learned, rocking it for many years long before the pandemic. There's quite a background about you having been a son of a homicide detective, private eye agency, and then you were going to become an FBI agent. And the quote from that in the article that's the Air Investigator is truly a classic. Yeah, you have in there, “I guarantee I'm the only public health student ever to fail an FBI lie detector polygraph in the morning and start graduate school a few hours later.” That's amazing. That's amazing.Joseph Allen (00:01:29):All right. Well, you've done your deep research apparently. That's good. Yeah, my dad was a homicide detective and I was a private investigator. That's no longer my secret. It's out in the world. And I switched careers and it happened to be the day I took the polygraph at the FBI headquarters in Boston, was the same day I started graduate studies in public health.Sick vs Healthy Buildings (Pre-Covid)Eric Topol (00:01:53):Well, you're still a detective and now you're a detective of everything that can hurt us or help us environmentally and my goodness, how grateful we are that you change your career path. I don't know anyone who's had more impact on buildings, on air, and we're going to get into chemicals as well. So if we go back a bit here, you wrote a book before the pandemic, talk about being prescient. It’s called Healthy Buildings: How Indoor Spaces Can Make You Sick - or Keep You Well with John Macomber, your co-author. What was it that gave you the insight to write a book before there was this thing called Covid?Joseph Allen (00:02:41):Yeah, well, thanks for making the connection too, my past career to current career. For many years, I thought there wasn't a connection, but I agree. There's actually a lot of similarities and I also am really appreciative. I am lucky I found the field of Public Health, it's clearly where I belong. I feel like I belong here. It's a place to make an impact that I want to make in my career. So yeah, the Healthy Buildings book, we started writing years before the pandemic and was largely motivated by, I think what you and others and other people in my field have known, is that buildings have an outsized impact on our health. Yet it's not something that comes to the forefront when you ask people about what matters for their health. Right, I often start presentations by asking people that, what constitutes healthy living? They'll say, I can't smoke, I have to eat well.(00:03:30):I have to exercise. Maybe they'll say, outdoor pollution’s bad for you. Very few people, if any, will say, well, the air I breathe inside my building matters a lot. And over the years I had started my public health career doing forensic investigations of sick buildings. People really can get sick in buildings. It can be anything from headaches and not being able to concentrate all the way to cancer clusters and people dying because of the building. And I've seen this in my career, and it was quite frustrating because I knew, we all knew how to design and operate buildings in a way that can actually keep people healthy. But I was frustrated like many in my field that it wasn't advancing. In other words, the science was there, but the practice wasn't changing. We were still doing things the wrong way around ventilation, materials we put in our building, and I would lecture over and over and give presentations and I decided I want to try something new.(00:04:22):I do peer-reviewed science. That's great. I write pieces like you for the public, and I thought we'd try a longer form piece in a book, and it's published by Harvard Press. John Macomber for those who know is a professor at Harvard Business School who's an expert in real estate finance. So he'd been talking about the economic benefits of healthier buildings and some hand waving as he describes around public health. I've been talking about the public health benefits and trying to wave an economic argument. We teamed up to kind of use both of our strengths to, I hope make a compelling case that buildings are good for health and they're also just good business. In other words, try to break down as many barriers as we can to adoption. And then the book was published right as Covid hit.Indoor Air Quality and CognitionEric Topol (00:05:05):Yeah. I mean, it's amazing. I know that typically you have to have a book almost a year ahead to have it in print. So you were way, way ahead of this virus. Now, I'm going to come back to it later, but there were two things beyond the book that are pretty striking about your work. One is that you did all these studies to show with people wearing sensors to show that when the levels of CO2 were high by sensors that their cognition indoors was suffering. Maybe you could just tell us a little bit about these sensors and why aren't we all wearing sensors so that we don't lose whatever cognitive power that we have?Joseph Allen (00:05:56):Well, yeah. First I think we will start having these air quality sensors. As you know, they're starting to become a lot more popular. But yeah, when I first joined the faculty full-time at Harvard, one of the first studies I conducted with my team was to look at how indoor air quality influences cognitive function. And we performed a double-blind study where we took people, office workers and put them in a typical office setting. And unbeknownst to them, we started changing the air they were breathing in really subtle ways during the day, so they didn't know what we were doing. At the end of the day, we administered an hour and a half long cognitive function battery, and like all studies, we control for things like caffeine intake, baseline cognitive performance, all the other factors we want to account for. And after controlling for those factors in a double-blind study, we see that indoor air quality, minor improvements to indoor air quality led to dramatic increases in cognitive function test scores across domains that people recognize as important for everyday life.(00:06:59):How do you seek out and utilize information? How do you make strategic decisions? How do you handle yourself during a crisis and importantly recover after that crisis? I don't mean the world's ending crisis. I mean something happens at work that's stressful. How do you handle that and how do you respond? Well, it turns out that amongst all the factors that influence how we respond there, indoor air quality matters a lot. We call that study the COGfx Study for cognitive function. We replicated it across the US, we replicated it across the world with office workers around the world, and again, always showing these links, the subtle impact of indoor air quality on cognitive function performance. Now, that also then starts to be the basis for some of the economic analysis we perform with my colleague at Harvard Business School. We say, well, look, if you perform this much better related to air quality, what would happen if we implemented this at scale in a business?(00:07:51):And we estimate that there are just massive economic gains to be had. On a per person basis, we found and published on this, that's about $6,000 to $7,000 per person per year benefit across a company. It could lead to 10% gains to the bottom line performance of the company. And again, I'm a public health professor. My goal is to improve people's health, but we add a lens, mental health, brain health is part of health, and we add the economic lens to say, look, this is good for a worker of productivity and the costs are downright trivial when you compare it against the benefits, even just including the cognitive function benefits, not even including the respiratory health benefit.Eric Topol (00:08:33):And I mean, it's so striking that you did these studies in a time before sensors were, and they still are not widely accepted, and it really helped prove, and when we start to fall asleep in a group session indoors, it may not just be because we didn't have enough sleep the night before, right.Joseph Allen (00:08:56):It's funny you say that. I talk about that too. It's like, do we actually need the study to tell us to quantify what we've all experienced these bad conference rooms, you get tired, you can't concentrate, you get sleepy while you're driving your car. Yeah, a whole bunch of other factors. Maybe the speaker's boring, but a key factor is clearly indoor air quality and things like good ventilation, the chemical load in the space are all contributing.Eric Topol (00:09:20):Yeah. No, it's pretty darn striking. Now we're going to get into the pandemic, and this of course is when your work finally crystallized that you've been working on this for years, and then finally your collaboration with some of the aerosol experts. It was a transdisciplinary synergy that was truly extraordinary. And when you were on 60 Minutes last October, you said, “Think about the public health gains we've made over the past hundred years. We've made improvements to water quality, outdoor air pollution, our food safety, we've made improvements to sanitation: absolute basics of public health. Where has indoor air been in that conversation?” You brought it to us. I mean, you led the Lancet Commission on this. You've done a White House Summit keynote. You had a lot of influence. Why did it take us to finally wake up to this issue that you've been working on for years?Covid is Airborne, DenialJoseph Allen (00:10:31):Yeah. Well, I appreciate that, but I also liked what you started with. I mean, there's been a lot of us pulling on this, and I think one of the magical moments, if you could say that when the pandemic happened was that it forced these collaborations and forced a lot of us in our field to be a bit more vocal. And even that comment about the gains we made in public health, that comes from an article that we co-authored with 40 plus scientists around the world in science, trying to drive home the point that we've ignored one of the key factors that determines our health. We were all frustrated at the beginning of the pandemic. The first piece I wrote was January 2020, talking about healthy buildings as the first line of defense, airborne spread, ventilation, filtration. I could not get it published. I could not get it published.(00:11:20):So I moved it to an international paper. I wrote it in the Financial Times in early February, but it wasn't until mid-March that the Times took my piece on this airborne spread buildings ventilation. At the same time, we know people like Linsey Marr, Rich Corsi, many others, Shelly Miller out there publishing, doing the fundamental research, all trying to elevate, and I think we started to find each other and say, hey, someone's trying to hit the medical journals. We're not landing there. I'm trying to hit the Times, and we’re not landing there. We're trying to get the reporters to pay attention. It's not landing there. Let's team up. Let's write these joint pieces. And I think what happened was you saw the benefit of the collective effort and interdisciplinary expertise, right? We could all start to come together, start instead of having these separate voices, a little bit of a unified voice despite important scientific minor disagreements, but start to say, hey, we started elevate each other and said, this is really important. It's the missing component of the messaging in the early days of the pandemic, and to know how to defend yourself.Eric Topol (00:12:20):Well, I think a lot of people think the big miss, and I know you agree, was the lack of recognition of aerosol transmission instead of just liquid droplets. But what you brought to this was really your priors on the buildings themselves and the ventilation systems and air quality that was highly, I mean, critical to it isn't just the aerosol, it's obviously how buildings are set up. Now, there's an amazing piece of course that appeared in the summer of 2021 called the Air Investigator, which profiled you, and in it brings up several things that finally are, we're starting to get our act together. I mean, ultimately there was in May 2023 years later, the CDC says, we're going to do something about this. Can you tell us what was this very distinct new path that the CDC was at least saying? And also couple that with whatever action if or not action has been taken.Joseph Allen (00:13:33):Yeah. So there really was a monumental shift that took, it was years in development, but we finally won the argument, collectively that airborne spread was the dominant mode of transmission. Okay, we got that. Then the question is, well, what changes? Do we actually get guidance here? And that took a little bit longer. I give Rochelle Walensky a lot of credit when she came into the CDC, we talked with her about this. That's when you start to first see ventilation starts showing up and the guidance, including guidance for schools. So I think that was a big win, but still no one was willing to set an official target or standard around higher ventilation rates. So that's important. Early in the pandemic, some people started to hear a message, yes, ventilation is important. What's the obvious next question, well, how much, what do I need? So in the summer of 2020, actually Shelly Miller and I collaborated on this.(00:14:23):We published some guidance on ventilation targets for schools. We said four to six air changes per hour (ACH) and target that. Well, it wasn't until 2023, spring of 2023 that you mentioned that CDC published target ventilation rates, and they went with five air changes per hour, which is right where we were talking about in summer 2020. It's what the Lancet of COVID-19 Commission adopted, but it's momentous in this way. It's the first time in CDCs history they've ever published a ventilation rate target for health. Now, I know this seems slow at the time, and it was, but if we think about some of the permanent gains that will come out of the pandemic. Pandemic changes society and science and policy and practice this, we are never going back. Now buildings will be a first line of defense for respiratory pathogens going forward that can no longer be ignored. And now we have the published target by CDC. That's a big deal beca…

Full show notes at the publisher

Steve Horvath: Our Epigenetic Age Clocks Aug 23, 2024

Show notes

Steve Horvath made the seminal discovery of the—Horvath Clock— an epigenetic clock based on DNA methylation, which is now being used extensively in medical research and offered commercially for individuals (←we talk about that!). He was on the faculty at UCLA from 2000-2022 as a Professor of Human Genetics and Biostatistics, and now works on anti-aging research at Altos Labs.A perspective on the importance of epigenetic clocks this week’s Nature”This insight is crucial for deriving reliable biological markers of ageing in tissues or blood. Such a feat has been accomplished through the ingenious identification of epigenetic clocks in our genome. But these insights are even more important for revealing targets that enable intervention in the ageing process.”A video snippet on vegetable intake and epigenetic clocks. Full videos of all Ground Truths podcasts can be seen on YouTube here. The audios are also available on Apple and Spotify.Transcript with links to Audio and External LinksEric Topol (00:06):Hello, it's Eric Topol with Ground Truths, and I've got a terrific guest with me today, Steve Horvath. He's a geneticist, a statistician, a mathematician. He's got a lot of background that has led to what is a landmark finding in biomedicine, the Horvath clock. So Steve, welcome.Steve Horvath (00:30):Thank you for having me.Eric Topol (00:33):Well, it's really fascinating. I followed your work for well over a decade since you introduced the pan-tissue clock in 2013, and it's fascinating to go back a bit on that finding, which initially, I guess was in saliva a couple of years prior, and then you found it everywhere you looked, wherever cells had a nucleus and tissues. And what gave you the sense that these markers of methylation on the DNA would give us some clues about the aging process? How did you even come about to make this discovery?SerendipitySteve Horvath (01:17):It was an accidental discovery because before the methylation clock, I had worked very hard on a gene expression clock, a transcriptomic biomarker. I mean, I was at the height of my energy levels. I worked really on weekends, really eight hour days during the week. But all the weekends I had collected a large set of gene expression data and I dredged the data. And for two years and I couldn't get anywhere, there was nothing I could do. But nowadays, of course, you see various publications where people built transcriptomic clocks. But back in the day when we had these arrays, I just couldn't see a signal. And then at some point I got roped into a study of homosexuality where my collaborator at UCLA wanted to see whether there's an epigenetic correlate of sexual orientation in saliva. And so yeah, being a biostatistician, I said, sure, I analyzed the data and I couldn't find any signal for homosexuality.(02:48):But then I just looked for an aging signal in the same, and really within an hour of analyzing the data, I knew that I have to completely drop gene expression. I need to go after methylation. And the signal is so profound, and as you said initially we looked at saliva samples and we thought, isn't it curious? You spit in a cup and you can measure someone's age. And we were of course, hoping that this could become a valuable readout of biologic age, but it took, of course, many years to realize that potential. Nowadays, there's several companies that offer a saliva based methylation clock test. But yeah, many years passed, and it was important to fill in the details and to build the case that methylation clocks are predictive of things we care about time to death or time to various forms of morbidity. So it took many, many years to analyze large cohort studies and to accumulate the evidence that it actually works.Eric Topol (04:16):Yeah, I mean, it was pretty amazing back almost a decade ago when I would see, we would take tissue or blood sample and look at your clock and it would say, age of the person is 75 years. And then we look at the actual age of the person who is 75 years to say, wait a minute, how can this be? So I mean, the plausibility of this discovery, if you look back, I mean you say, well, this is just kind of the rust of the pipes, or how do you process that the methylation is such a marker potentially of a person's biologic age? Of course, we're going to get into how it could be a way to intervene to change the aging process. But would it be fair to say that its epigenetic clocks are not the same as biologic aging or how do you put all that together?Epigenetic Age vs Biologic AgeSteve Horvath (05:21):Yes, for sure. An epigenetic age estimate is certainly not the same as a biologic age estimate. And the reason why I say it is because biologic age is really determined by so many things and by so many organs. And as I mentioned initially, we had a clock for saliva later for blood and so on. And so, if you only have an epigenetic readout of a certain cell type, it's really too limited to assess the whole organismal state. And arguably you would want to measure also proteomics, readouts and many other data modalities. So I typically avoid the terminology biologic age, because to begin with, we don't have a definition of it. Decades of discussions, nobody really has a precise definition of it.Second Generation Epigenetic ClocksEric Topol (06:35):Well, from the first generation Horvath clock then became this newer second generation, GrimAge, PhenoAge, the DunedinPACE of aging. How has that helped to advance the field? Because as you touched on, they're measuring different things and what is it meant by kind of a second generation clock?Steve Horvath (07:03):Yeah, so a second generation clock truly aims to predict mortality or morbidity risk. As opposed to simply chronologic age or what is known as calendar age. And fortunately, there's no doubt that the second generation clocks can do that. I often finish a talk on GrimAge by telling the audience that I give them a money back guarantee, that it will be predictive of mortality in their cohort study. I'm 100% certain that it works if you analyze a hundred people or so. The question is more whether an individual could benefit from such a test. And there are now many providers of various epigenetic clock tests. These biomarkers have different names, but they're quite pricey. A couple of hundred dollars are needed to get such a measurement. And the question is, is it helpful for the individual should you get such a test? And I would say we are not quite there yet for a variety of reasons. The main reason being we don't have good interventions against accelerated epigenetic age. So because when you think about it, why does a doctor order a test for you? For example, cholesterol levels. Well, because they have a drug against elevated cholesterol levels, the statin. And at the moment, we don't have validated interventions against accelerated epigenetic age. So that's kind of missing.Eric Topol (09:13):Yeah, we're going to get to that because obviously a lot of things are in the pipeline there, but are you saying then that these people that are getting these consumer tests, that they're getting a test that really wasn't validated at an individual level, so it predicts their mortality that it may be good at a cohort or population level, but maybe it's not so helpful, accurate, or would you say it is accurate? I mean, GrimAge is a good name because since it says when you're going to die. How do you make the differentiation between the individual level or beyond?Steve Horvath (09:59):Yeah, I think it's good to compare to other biomarkers. So take glucose levels, hemoglobin A1C, nobody doubts that these levels predict mortality risk when you study couples a hundred people. But how accurate is such a test for an individual? Clearly there is substantial noise associated with a prediction. Two people could have exactly the same hemoglobin A1C levels, but live very different lifespans. And the same holds for epigenetic clocks. They do predict how long you live. In theory, one could arrive at an estimate of age and death. There's a complicated mathematical formula that allows you to do that, but there would be a substantial error bar associated with it, an order of magnitude plus minus five years. And so, for the individual, such an estimate is not that important because the error bar is substantial. But I want to add that these second generation clocks, they do predict mortality risk. There's no question.Maximal LifespanEric Topol (11:35):Well, as you know, the longevity space is now very crowded with all sorts of clubs, and it's like a circus out there. And some of these things are being promoted that really don't have the basis or have a false sense to consumers who want to live forever and be healthy forever. But maybe these markers are not really helping guide them so much. Now, you recently published you and your group a fascinating paper, so getting away from the individual for a second, but now at the species level and in Science Advances, and we'll put this diagram with the podcast, but you looked at 348 mammal species for the maximal lifespan with DNA methylation. And it was amazing to see the display from the desert hamster all the way to the humpback whale with somewhere along the way, the humans. So you could predict maximal lifespan pretty well, right?Steve Horvath (12:43):Yes. So I collected this very large dataset over seven years, and one of the reasons was to understand the mystery of maximum lifespan. The bowhead whale can live over 211 years, whereas certain mice only three or four years. And my question was, can methylation teach us something about maximum lifespan? And the answer is a resounding, yes. The methylation profiles very much predict the maximum lifespan of a species. And maybe to use a metaphor to explain the patterns. So one can visualize methylation around the DNA molecule, like a landscape. You want that certain regions exhibit high levels of methylation. These regions must be really shut down and other parts of the DNA as opposed to exhibit very low methylation, for example, a transcriptional start sites. And long lived species have a very hilly landscapes, high hills of methylation and steep valleys of low methylation. Where shorter lived species have flatter landscapes. So that was one of the insights of that study. The other perhaps paradoxical insight was that the locations in our DNA that gain methylation with chronologic age, these regions often differ from regions that determine the maximum lifespan of our species. So that's a bit perhaps paradoxical and counterintuitive, but it just shows that the DNA encodes our species characteristics at different locations from our mortality risk.The Other ClocksEric Topol (15:13):Right. No, and I mean it's fascinating. I can imagine how it could take seven years to pull all that data together. It's amazing. Now, one of the issues of course, is if you're trying to gauge the biologic age, which we already established is somewhat different than epigenetic age or a clock, there are many different ways to do that. And you mentioned transcriptome clocks, which are not as well perhaps developed. Obviously, none of these others are developed like the Horvath clock and newer generation clocks, but there's immuno aging clocks like iAge, there's proteomic clocks, there's organ clocks with high-throughput proteomics, thousands of proteins. Do you see these as complimentary, like orthogonal where they each add to the story? Or do you really see the methylation as distinct?Steve Horvath (16:20):Well, I think ideally you measure all of the above to really get a very granular understanding of different facets of aging. And however, scientists always like to find deep connections between different readouts. For example, it would be wonderful if we could use proteomics instead of methylation, or my group has worked on the opposite. So we can actually estimate protein levels in the plasma based on methylation for about 10% of all plasma proteins, you can estimate their levels based on methylation. So yeah, people who are interested in these deeper programs that ideally link everything, some sort of aging program that underlies these different manifestations of aging, they will want to reduce everything. But until we have a deeper understanding, I think let's air on the side of measuring too much.Eric Topol (17:45):Well, what's interesting, as you mentioned, I didn't realize you could basically impute the protein story from the methylation, but one of the issues is if you want to do 11,000 plasma proteins, it could cost a thousand dollars. But if you want to do a bisulfite methylation, you might do that for very inexpensively. So there's a practical part of this too, and the immune characterization is even more expensive and difficult from a practical standpoint. So we go back to that initial work that you did and how you got into an area that is practical, inexpensive compared to some of the alternatives. But as you say, they may have features that are also helpful. Now, this is now the craze, this epigenetic clocks, and I want to mention you probably didn't see it because it's not a journal that you would look at, but just yesterday, July 29th, there were 12 papers published in JAMA Network Open.Modulating Your Epigenetic Clock(18:51):Everything from how loss of loved ones changes your epigenetic clock to PTSD, to vegan diets, to inequities. I mean, just incredible. So it is the rage now. It's taken the biomedical community some years to catch up to where you were. And one of the things of course that we know that from your prior work that is an intervention that helps give a less accelerated epigenetic clock is exercise. And in fact, that was highlighted in our Lancet essay in the first week August issue. But can you comment on that and anything else that we know like plant diets and anything that favorably influence our DNA methylation pattern?Steve Horvath (19:52):Yes. So interestingly, vegetable intake really has a strong effect on GrimAge and many other epigenetic clocks. And maybe this is obvious to the listener, everybody knows that vegetable intake is healthy. However, it's very surprising to me as a scientist to contemplate how is it that vegetable intake affects the methylation levels of your blood? How does it affect the hematopoietic stem cells? I just don't understand the mechanism behind it, and however, the effect is very strong. So we studied postmenopausal women in the women's health initiative, and for these women, we had blood measures of carotenoid levels. So this is an objective measure of vegetable intake, and the correlations were substantial. So that's one intervention I'm quite certain about. Other intervention that have a strong effect relate to metabolic syndrome, anything that relates to type 2 diabetes such as obesity, high glucose levels, that part of the biology very much affects our epigenetic clocks. So disturbed metabolism has a strong effect.Eric Topol (21:37):Has these findings changed your diet or made you exercise more or anything like that?Steve Horvath (21:44):. So I eat a lot of frozen vegetables. My freezer as full frozen vegetables.Eric Topol (21:56):That's great. Well, there's a lot of uses today as we touched on in the Lancet piece as we're waiting for more benchmarking and more work on this. But for example, we have a shortage of donor organs, and there are people who might be of calendar age advanced, but their epigenetic clock might put them at a much younger age. Is that ready for use in the transplant world as one application?Steve Horvath (22:37):I haven't seen that yet. I've seen…

Full show notes at the publisher

Pradeep Natarajan: Preventing Heart Disease Aug 13, 2024

Show notes

Pradeep is a brilliant geneticist and Director of Preventive Cardiology, holds the Paul & Phyllis Fireman Endowed Chair in Vascular Medicine at Mass General Hospital and on faculty at Harvard Medical School and the Broad Institute. His prolific research has been illuminating for the field of improving our approach to reduce the risk of heart disease. That’s especially important because heart disease is the global (and US) #1 killer and is on the increase. We didn’t get into lifestyle factors here since there was so much ground to cover on new tests. drugs, and strategies.A video snippet of our conversation on ApoB. Full videos of all Ground Truths podcasts can be seen on YouTube here. The audios are also available on Apple and Spotify.Transcript with links to key publications and audioEric Topol (00:06):Well, welcome to Ground Truths. I'm Eric Topol and with me is Pradeep Natarajan from Harvard. He's Director of Preventative Cardiology at the Mass General Brigham Health System and he has been lighting it up on the field of cardiovascular. We're going to get to lots of different parts of that story and so, Pradeep welcome.Pradeep Natarajan (00:31):Thanks Eric, really delighted and honored to be with you and have this discussion.Eric Topol (00:36):Well, for years I've been admiring your work and it's just accelerating and so there's so many things to get to. I thought maybe what we'd start off with is you recently wrote a New England Journal piece about two trials, two different drugs that could change the landscape of cardiovascular prevention in the future. I mean, that's one of the themes we're going to get to today is all these different markers and drugs that will change cardiology as we know it now. So maybe you could just give us a skinny on that New England Journal piece.Two New Lipid Targets With RNA DrugsPradeep Natarajan (01:16):Yeah, yeah, so these two agents, the trials were published at the same time. These phase two clinical trials for plozasiran, which is an siRNA against APOC3 and zodasiran, which is an siRNA against ANGPTL3. The reason why we have medicines against those targets are based on human genetics observations, that individuals with loss of function mutations and either of those genes have reduced lipids. For APOC3, it's reduced triglycerides for ANGPTL3 reduced LDL cholesterol and reduced triglycerides and also individuals that have those loss of function mutations also have lower risk for coronary artery disease. Now that's a very similar parallel to PCSK9. We have successful medicines that treat that target because people have found that carriers of loss of function mutations in PCSK9 lead to lower LDL cholesterol and lower coronary artery disease.(02:11):Now that suggests that therapeutic manipulation without significant side effects from the agents themselves for APOC3 and ANGPTL3 would be anticipated to also lower coronary artery disease risk potentially in complementary pathways to PCSK9. The interesting thing with those observations is that they all came from rare loss of function mutations that are enriched in populations of individuals. However, at least for PCSK9, has been demonstrated to have efficacy in large groups of individuals across different communities. So the theme of that piece was really just the need to study diverse populations because those insights are not always predictable about which communities are going to have those loss of function mutations and when you find them, they often have profound insights across much larger groups of individuals.Eric Topol (03:02):Well, there's a lot there that we can unpack a bit of it. One of them is the use of small interfering RNAs (siRNA) as drugs. We saw in the field of PCSK9, as you mentioned. First there were monoclonal antibodies directed against this target and then more recently, there’s inclisiran which isn't an RNA play if you will, where you only have to take it twice a year and supposedly it's less expensive and I’m still having trouble in my practice getting patients covered on their insurance even though it's cheaper and much more convenient. But nonetheless, now we're seeing these RNA drugs and maybe you could comment about that part and then also the surprise that perhaps is unexplained is the glucose elevation.Pradeep Natarajan (03:53):Yeah, so for medicines and targets that have been discovered through human genetics, those I think are attractive for genetic-based therapies and longer interval dosing for the therapies, which is what siRNAs allow you to do because the individuals that have these perturbations, basically the naturally occurring loss of function mutations, they have these lifelong, so basically have had a one-time therapy and have lived, and so far, at least for these targets, have not had untoward side effects or untoward phenotypic consequences and only reduce lipids and reduce coronary artery disease. And so, instead of taking a pill daily, if we have conviction that that long amount of suppression may be beneficial, then longer interval dosing and not worrying about the pill burden is very attractive specifically for those specific therapeutics. And as you know, people continue to innovate on further prolonging as it relates to PCSK9.(04:57):Separately, some folks are also developing pills because many people do feel that there's still a market and comfort for daily pills. Now interestingly for the siRNA for zodasiran at the highest dose, actually for both of them at the highest doses, but particularly for zodasiran, there was an increase in insulin resistance parameters actually as it relates to hyperglycemia and less so as it relates to insulin resistance, that is not predicted based on the human genetics. Individuals with loss of function mutations do not have increased risks in hyperglycemia or type 2 diabetes, so that isolates it related to that specific platform or that specific technology. Now inclisiran, as you'd mentioned, Eric is out there. That's an siRNA against PCSK9 that's made by a different manufacturer. So far, the clinical trials have not shown hyperglycemia or type 2 diabetes as it relates inclisiran, so it may be related to the specific siRNAs that are used for those targets. That does merit further consideration. Now, the doses that the manufacturers do plan to use in the phase three clinical trials are at lower doses where there was not an increase in hyperglycemia, but that does merit further investigation to really understand why that's the case. Is that an expected generalized effect for siRNAs? Is it related to siRNAs for this specific target or is it just related to the platform used for these two agents which are made by the same manufacturer?Eric Topol (06:27):Right, and I think the fact that it's a mystery is intriguing at the least, and it may not come up at the doses that are used in the trials, but the fact that it did crop up at high doses is unexpected. Now that is part of a much bigger story is that up until now our armamentarium has been statins and ezetimibe to treat lipids, but it's rapidly expanding Lp(a), which for decades as a cardiologist we had nothing to offer. There may even be drugs to be able to lower people who are at high risk with high Lp(a). Maybe you could discuss that.What About Lp(a)?Pradeep Natarajan (07:13):Yeah, I mean, Eric, as you know, Lp(a) has been described as a cardiovascular disease risk factors for quite so many years and there are assays to detect lipoprotein(a) elevation and have been in widespread clinical practice increasing widespread clinical practice, but we don't yet have approved therapies. However, there is an abundance of literature preclinical data that suggests that it likely is a causal factor, meaning that if you lower lipoprotein(a) when elevated, you would reduce the risk related to lipoprotein(a). And a lot of this comes from similar human genetic studies. The major challenge of just relating a biomarker to an outcome is there are many different reasons why a biomarker might be elevated, and so if you detect a signal that correlates a biomarker, a concentration to a clinical outcome, it could be related to that biomarker, but it could be to the other reasons that the biomarker is elevated and sometimes it relates to the outcome itself.(08:10):Now human genetics is very attractive because if you find alleles that strongly relate to that exposure, you can test those alleles themselves with the clinical outcome. Now the allele assignment is established at birth. No other factor is going to change that assignment after conception, and so that provides a robust, strong causal test for that potential exposure in clinical outcome. Now, lipoprotein(a) is unique in that it is highly heritable and so there are lots of different alleles that relate to lipoprotein(a) and so in a well powered analysis can actually test the lipoprotein(a) SNPs with the clinical outcomes and similar to how there is a biomarker association with incident myocardial infarction and incident stroke, the SNPs related to lipoprotein(a) show the same. That is among the evidence that strongly supports that this might be causal. Now, fast forward to many years later, we have at least three phase three randomized clinical trials testing agents that have been shown to be very potent at lowering lipoprotein(a) that in the coming years we will know if that hypothesis is true. Importantly, we will have to understand what are the potential side effects of these medicines. There are antisense oligonucleotides and siRNAs that are primarily in investigation. Again, this is an example where there's a strong genetic observation, and so these genetic based longer interval dosing therapies may be attractive, but side effects will be a key thing as well too. Those things hard to anticipate really can anticipate based on the human genetics for off target effects, for example.(09:52):It's clearly a risk signal and hopefully in the near future we're going to have specific therapies.Eric Topol (09:57):Yeah, you did a great job of explaining Mendelian randomization and the fact the power of genetics, which we're going to get into deeper shortly, but the other point is that do you expect now that there's these multiple drugs that lower Lp(a) efficiently, would that be enough to get approval or will it have to be trials to demonstrate improved cardiovascular outcomes?Pradeep Natarajan (10:24):There is a great regulatory path at FDA for approval just for LDL cholesterol lowering and inclisiran is on the market and the phase three outcomes data has not yet been reported because there is a wide appreciation that LDL cholesterol lowering is a pretty good surrogate for cardiovascular disease risk lowering. The label will be restricted to LDL cholesterol lowering and then if demonstrated to have clinical outcomes, the label could be expanded. For other biomarkers including lipoprotein(a), even though we have strong conviction that it is likely a causal factor there hasn't met the bar yet to get approval just based on lipoprotein(a) lowering, and so we would need to see the outcomes effects and then we would also need to understand side effects. There is a body of literature of side effects for other therapies that have targeted using antisense oligonucleotides. We talked about potential side effects from some siRNA platforms and sometimes those effects could overtake potential benefits, so that really needs to be assessed and there is a literature and other examples.(11:31):The other thing I do want to note related to lipoprotein(a) is that the human genetics are modeled based on lifelong perturbations, really hard to understand what the effects are, how great of an effect there might be in different contexts, particularly when introduced in middle age. There's a lot of discussion about how high lipoprotein(a) should be to deliver these therapies because the conventional teaching is that one in five individuals has high lipoprotein(a), and that's basically greater than 75 nanomoles per liter. However, some studies some human genetic studies to say if you want to get an effect that is similar to the LDL cholesterol lowering medicines on the market, you need to start with actually higher lipoprotein(a) because you need larger amounts of lipoprotein(a) lowering. Those are studies and approaches that haven't been well validated. We don't know if that's a valid approach because that's modeling based on this sort of lifelong effect. So I'm very curious to see what the overall effect will be because to get approval, I think you need to demonstrate safety and efficacy, but most importantly, these manufacturers and we as clinicians are trying to find viable therapies in the market that it won't be hard for us to get approval because hopefully the clinical trial will have said this is the context where it works. It works really well and it works really well on top of the existing therapies, so there are multiple hurdles to actually getting it directly to our patients.How Low Do You Go with LDL Cholesterol?Eric Topol (13:02):Yeah, no question about that. I'm glad you've emphasized that. Just as you've emphasized the incredible lessons from the genetics of people that have helped guide this renaissance to better drugs to prevent cardiovascular disease. LDL, which is perhaps the most impressive surrogate in medicine, a lab test that you already touched on, one of the biggest questions is how low do you go? That is Eugene Braunwald, who we all know and love. They're in Boston. The last time I got together with him, he was getting his LDL down to close to zero with various tactics that might be extreme. But before we leave these markers, you're running preventive cardiology at man's greatest hospital. Could you tell us what is your recipe for how aggressive do you go with LDL?Pradeep Natarajan (14:04):Yeah, so when I talk to patients where we're newly getting lipid lowering therapies on, especially because many people don't have a readout of abnormal LDL cholesterol when we're prescribing these medicines, it's just giving them a sense of what we think an optimal LDL cholesterol might be. And a lot of this is based on just empirical observations. So one, the average LDL cholesterol in the modern human is about 100 to 110 mg/dL. However, if you look at contemporary hunter gatherers and non-human primates, their average LDL is about 40 to 50 and newborn babies have an LDL cholesterol of about 30. And the reason why people keep making LDL cholesterol lowering medicines because as you stack on therapies, cardiovascular disease events continue to reduce including down to these very low LDL cholesterol values. So the population mean for LDL cholesterol is high and everybody likely has hypercholesterolemia, and that's because over the last 10,000 years how we live our lives is so dramatically different and there has not been substantial evolution over that time to change many of these features related to metabolism.(15:16):And so, to achieve those really low LDL cholesterol values in today's society is almost impossible without pharmacotherapies. You could say, okay, maybe everybody should be on pharmacotherapies, and I think if you did that, you probably would reduce a lot of events. You'll also be treating a lot of individuals who likely would not get events. Cardiovascular disease is the leading killer, but there are many things that people suffer from and most of the times it still is not cardiovascular disease. So our practice is still rooted in better identifying the individuals who are at risk for cardiovascular disease. And so, far we target our therapies primarily in those who have already developed cardiovascula…

Full show notes at the publisher

Shane Crotty: A Landmark Study on Upper Airway Mucosal Immunity Jul 31, 2024

Show notes

A video snippet of our conversation. Full videos of all Ground Truths podcasts can be seen on YouTube here. The audios are also available on Apple and Spotify.Shane Crotty: A Landmark Study on Upper Airway Mucosal ImmunityTranscriptThis is the first time a Ground Truths podcast is being posted simultaneous with a new publication, this one in Nature, by Professor Shane Crotty and his colleagues at La Jolla Institute for Immunology. Shane is one of the leading immunologists and virologists in the country; he and his group published in 2020 the first detailed analysis for how our immune system responds to SARS-CoV-2. Shane also, among many other notable contributions during COVID, illuminated the role of hybrid immunity vs COVID, the differences between and additivity of vaccination and infection.Today’s paper in Nature is indeed a landmark contribution doing something that hasn’t been done before—to understand the underpinnings of mucosal immunity of the upper airway. 100 participants had monthly nasal and nasopharyngeal swabs throughout the pandemic. With a median of >100,000 cells per swab recovered, they undertook single-cell sequencing and full characterization of the cells (tissue-resident memory B cells, CD4+ and CD8+ T cells, germinal center follicular helper T cells and B cells, etc.) to determine optimal immune protection of the upper airway, the effect of infections by different variants, breakthrough infections, vaccination, and age.Here is the transcript of our conversation about the new report with links to the audio:Eric Topol (00:06):Hello, it's Eric Topol with Ground Truths, and with me today is Professor Shane Crotty from the La Jolla Institute of Immunology (LJI), not too far away from where I work at Scripps. And Shane has been a go-to immunologist colleague here in the Mesa, and he and his colleagues were the ones that really first published the response to SARS-CoV-2 as far as the immunologic response. And today we're doing something very unique. We're going to go over for the first time in the two year plus history of Ground Truths, going to have a publication with at least simultaneous or near simultaneous podcast. Shane, welcome and congratulations on this really important paper in Nature.Shane Crotty (00:57):Thanks, Eric. Thanks for having me. Yeah, somebody asked if I was going to go over to Scripps for the podcast and I was like, yeah, we could.Eric Topol (01:06):You could. You could. But no, it's good. And it's nice having the logo of this great institute you work at right in the right corner. And you've done so many contributions with your colleagues at La Jolla Institute. It's really a privilege to have a chance to learn from you and particularly about what we're going to talk about today, which is mucosal immunity to upper airway infections, which is especially germane to COVID. And we're actually in the middle of a significant wave of COVID right now. And I guess it would maybe be fair to say, Shane, that we've never truly understood the underpinnings, the real details of upper airway mucosal immunity. Is that a fair statement?Shane Crotty (01:53):Yeah, it is a fair statement.Eric Topol (01:56):Okay. So today we're going to crack the case. This paper from you and your colleagues, of course, you're the senior author and first author, Sydney Ramirez did a remarkable study. I mean, just extraordinary. This is why we're doing a special podcast about it. Maybe you could just kind of give us the overview of the design because you were doing things that haven't been done before.Shane Crotty (02:24):Sure. And, I would say the genesis even of it goes back to what you were introducing. I mean, during the pandemic, we like a lot of scientists spent a lot of time and energy trying to help understanding immune responses to this virus, and immune memory to this virus, and what was involved in protective immunity. And we're certainly proud of the work that we did. And it was hard work. And after a while we were exhausted and we stopped.Shane Crotty (02:59):And then we came back to it after a while and said, well, the virus is still here. And so many people have contributed so much to better understanding the virus and creating vaccines. But there are clearly still things we don't understand. What are those biggest knowledge gaps and where might we be able to contribute? And really to me the biggest one was location, location, location. This is a virus that infects your nose, infects your upper airway—your nose, and throat, and oral cavity. And then obviously if you get severe disease, the severe disease and death are from the lungs. And it's just been a big knowledge gap in terms of understanding what actually occurs in those tissues immunologically and what is associated with protective immunity or what could be associated with protective immunity. And sort of looking forward what might be helpful for mucosal vaccine development from things that we could learn.Shane Crotty (04:12):So we started from what we would call the basics, and what does immune memory look like in the upper airways in normal people? And that hasn't been available really even in, and we started this two years ago, even in the biggest atlases published of the human body. There was no upper airway tissue representation at all. And that's because technically it's just tough to access and difficult to reproducibly get at. And so, we recruited people to a group of 20 to 30 people to come to LJI once a month, and just started testing out, published and unpublished sampling techniques to see were there ways where we could reproducibly sample immune cells in the upper airways from people. And once we got things, so the keys for us were you got to have enough cells that you can collect to learn something from. And luckily with modern techniques of flow cytometry and single cell sequencing, you don't need that many cells. And so, we could get a hundred thousand cells on a swab and that's enough to do a lot with. And second, how reproducible was it? So we showed, we had people come in every month for a year and we could reproducibly find the same things in their swab; same cell types in their swabs. And the third thing was that people would come back.Shane Crotty (06:05):We found that if you have good nurses doing the techniques, we could find ways that this would be a sampling approach that was tolerable and people would come back for repeat measures, which is really valuable to see what's happening in people over time. So that was what we started from in the study and built from.Eric Topol (06:27):And if I am correct, you sampled two places with the swabs, one in the nose and one of the throat. Or, I think one which you have in the paper as the MT for something about the median nasal turbinate and the other adenoid in the back of the throat. Is that right?Shane Crotty (06:50):So all the sampling is a swab into your nose. And when we were doing that, we were really excited to see the diversity of immune cells, particularly T cells and B cells, memory T cells and B cells that we isolated. They're like, wow, there's actually a lot of interesting immune memory up in there. And the lab said, oh, by the way, we're seeing T follicular helper cells (TFH). Now that happens to be my favorite cell type.Eric Topol (07:22):Why is that, Shane? Of all the cells, why do you say that's your favorite? I know you publish a lot on it.Shane Crotty (07:31):Because those are the T cells that are required for basically all neutralizing antibody responses. All high-quality antibody responses depend on—almost all high-quality antibody responses depend on—T cell help. That T cell help comes from T follicular helper cells. Antibody evolution is certainly one of the coolest processes of the immune system. And all of that depends on T follicular helper cells. So the fact that for example, you could get Omicron neutralizing antibodies even after only being vaccinated with ancestral vaccine, that's the immune system making guesses of what variants would look like. And those guesses come about through this antibody evolution that's driven by T follicular helper cells. So, it's really one of the most brilliant things the immune system does, and that's a cell type that's really key, but those processes happen in lymphoid tissue. That's what happens in lymph nodes and spleen. And here we were sampling epithelium, your nasal epithelium, so the cells didn't really belong there.Shane Crotty (08:37):And so, that's what turned the study in another direction. And we said, okay, let's figure out why is it that these cells are present in these swabs? And we had a couple of possibilities. One possibility was that the swab was going all the way back to the posterior wall of your nasopharynx, your top of your throat and sampling adenoid tissue. So adenoid tonsils and adenoids are a true lymphoid tissue and they're a mucosal lymphoid tissue. And so, we came up with multiple ways to validate that that's what we were testing. And in fact, it was the Sydney Ramirez, a clinician, and the ENTs involved who said, well, let's just look. And so, they actually did endoscopies with the swab to actually see where the swab went. We've got videos of the swabs going into the adenoid crypt in the back, and then we've got measurements of here are the cells that you find on those swabs.Shane Crotty (09:58):And what's cool about it is that, yes, so we did studies with two sets. We then shifted to doing studies with two sets of swabs. One where we essentially went “halfway back” where we were detecting that epithelium of your nasal passages and then one where it was all the way back and detecting the adenoid lymphoid tissue. So here we've got two different sites in your upper airways that are about an inch apart, and we're detecting essentially completely different cells of the immune system at those two places. And we tend to think of the cells present in that epithelial tissue as probably the sentinels, the cells that are sitting there that can potentially immediately react and try and protect you against a viral or bacterial infection. Whereas the lymphoid tissue, the adenoids, is really about generating the immune responses in the first place and priming immune responses. And that's where these germinal centers can occur, which are where the TFH are where you can get antibody evolution. And so, we found in the course of the study that with this non-invasive technique that we can.Eric Topol (11:14):By the way, I don't want to be signing up for the one way up there because I mean just a mid-nose enough for me. So wow, I got to give credit to your study participants for coming back every month for a year to have that. Some people call it a brain biopsy.Video of swab of nasopharyngeal tissueShane Crotty (11:33):Right. So I will tell you, it is a different experience than the COVID nasopharyngeal swab might've gotten through your car window. If you're actually sitting down in a comfortable space and there's a nurse doing it with these particular goals. We really found, we had a hundred people in the study and a total of 300 swabs, and the vast majority of people came back if we asked them to.Eric Topol (12:06):That's great.Shane Crotty (12:07):And we're certainly very thankful for the volunteers. Obviously they were volunteering in the first place to participate. So I'm a little hesitant about the video because I've told people to not show it to potential volunteers because it definitely doesn't encourage you to volunteer. You're like, wait, that's what's happening? But actually, I've had it done on me.Video of the swab to the nasopharynx for adenoid (lymphoid tissue) access.Eric Topol (12:37):Not that bad.Shane Crotty (12:39):It's really pretty compelling. And by doing these repeated samples, we actually now have the capacity to look at ongoing immune responses like after an infection or vaccination in people and see how that results in the immune system changing and what might be the source of the protective immunity that comes up. So we've actually got data in the paper looking at this antibody evolution in real time. So we've got affinity maturation of B cells occurring in just normal healthy adults of mucosal B cells against COVID. And so, that's really helping us learn what's possible, basically to figure out, okay, if you're going to try and make a vaccine, what types of immune cells are even possible to generate in this tissue? And where might you try and generate them? Or if you're trying to study some disease state, what are types of cells that might be problematic?Eric Topol (13:45):Yeah, I mean, I think the idea that so many of us have been pushing for a nasal vaccine to induce mucosal immunity because, as you know very well, the current shots are not very good at any durable or substantial protection from upper airway infections of COVID or SARS-CoV-2 and other infections. So I think one of the most important parts of this report is that it lends itself well to helping towards artificially, if you will, make a vaccine to get the protective features that you were able to identify. Maybe you could just [speculate], if you had the ideal nasal airway, what would the cellular profile look like?Shane Crotty (14:44):Ah, I see. Yeah, great question. So, first of all, antibodies are great. So most of my career has been dedicated to most licensed vaccines. The correlate of protection is antibodies. Antibodies clearly can be protective, and if you can get them that’s excellent, so certainly I would want, in terms of the non-cellular component, I would want antibodies present, neutralizing antibodies present in it.Eric Topol (15:26):Are these IgA or IgG?Shane Crotty (15:31):Yeah, in an ideal situation, what would I want? I'd want a mix of both, basically. The IgAs look like they have a little more protective efficacy, but the IgGs, just at a molecular level have a longer half-life, stick around a little. So yeah, I'd want both. And then really the premise for most of what we do is saying, in situations where antibody isn't enough or the antibodies don't stay around long enough, or you've got a variant that now obviates the protective efficacy of that particular antibody, are there other types of protective immunity you can have? And the immune system has other stuff besides antibodies for a reason. Of the lymphocytes in your blood, most of them aren't antibody producing cells. Most of them are other things. And so, well sticking with adjacent to antibodies, those antibodies in the mucosa, I'd want them to be made by cells that were literally right there. So plasma cells living in that site so that you've got basically the highest concentration of antibodies you can get because they're not having to diffuse through the whole body. They're just already at their highest concentration right there. Now antibodies come from B cells, that's what encodes the antibodies.Shane Crotty (17:03):And so, the B cells can make neutralizing antibodies if it turns out that you haven't made enough neutralizing antibodies, or if there's a variant that escapes those, maybe there are other B cells that could make, once you get infected, more B cells that could make more antibody rapidly infection, or B cells that recognize this variant that is mismatched to the current antibodies you have. But memory B cells are basically a library of different antibody specificities representing different guesses about what viral variants or structures might look like. And so, I would want memory B cells in that upper airway tissue…

Full show notes at the publisher

Faisal Mahmood: A.I.'s Transformation of Pathology Jul 28, 2024

Show notes

Full videos of all Ground Truths podcasts can be seen on YouTube here. The audios are also available on Apple and Spotify.Thank you for reading Ground Truths. This post is public so feel free to share it.Transcript with audio and external linksEric Topol (00:05):Hello, it's Eric Topol with Ground Truths, and I am really thrilled to have with me Professor Faisal Mahmood, who is lighting it up in the field of pathology with AI. He is on the faculty at Harvard Medical School, also a pathologist at Mass General Brigham and with the Broad Institute, and he has been publishing at a pace that I just can't believe we're going to review that in chronological order. So welcome, Faisal.Faisal Mahmood (00:37):Thanks so much for having me, Eric. I do want to mention I'm not a pathologist. My background is in biomedical imaging and computer science. But yeah, I work very closely with pathologists, both at Mass General and at the Brigham.Eric Topol (00:51):Okay. Well, you know so much about pathology. I just assume that you were actually, but you are taking computational biology to new levels and you're in the pathology department at Harvard, I take it, right?Faisal Mahmood (01:08):Yeah, I'm at the pathology department at Mass General Brigham. So the two hospitals are now integrated, so I'm at the joint department.Eric Topol (01:19):Good. Okay. Well, I'm glad to clarify that because as far as I knew you were hardcore pathologist, so you're changing the field in a way that is quite unique, I should say, because a number of years ago, deep learning was starting to get applied to pathology just like it was and radiology and ophthalmology. And we saw some early studies with deep learning whereby you could find so much more on a slide that otherwise would be not even looked at or considered or even that humans wouldn't be able to see. So maybe you could just take us back first to the deep learning phase before these foundation models that you've been building, just to give us a flavor for what was the warmup in this field?Faisal Mahmood (02:13):Yeah, so I think around 2016 and 2017, it was very clear to the computer vision community that deep learning was really the state of the art where you could have abstract feature representations that were rich enough to solve some of these fundamental classification problems in conventional vision. And that's around the time when deep learning started to be applied to everything in medicine, including pathology. So we saw some earlier cities in 2016 and 2017, mostly in machine learning conferences, applying this to very basic patch level pathology dataset. So then in 2018 and 2019, there were some studies in major journals including in Nature Medicine, showing that you could take large amounts of pathology data and classify what's known to us and including predicting what's now commonly referred to as non-human identifiable features where you could take a label and this could come from molecular data, other kinds of data like treatment response and so forth, and use that label to classify these images as responders versus non-responders or having a certain kind of mutation or not.(03:34):And what that does is that if there is a morphologic signal within the image, it would pick up on that morphologic signal even though humans may not have picked up on it. So it was a very exciting time of developing all of these supervised, supervised foundation models. And then I started working in this area around 2019, and one of the first studies we did was to try to see if we can make this a little bit more data efficient. And that's the CLAM method that we published in 2021. And then we took that method and applied it to the problem of cancers of unknown primary, that was also in 2021.Eric Topol (04:17):So just to review, in the phase of deep learning, which was largely we're talking about supervised with ground truth images, there already was a sign that you could pick up things like the driver mutation, the prognosis of the patient from the slide, you could structural variations, the origin of the tumor, things that would never have been conceived as a pathologist. Now with that, I guess the question is, was all this confined to whole slide imaging or could you somehow take an H&E slide conventional slide and be able to do these things without having to have a whole slide image?Faisal Mahmood (05:05):So at the time, most of the work was done on slides that were fully digital. So taking a slide and then digitizing the image and creating a whole slide image. But we did show in 2021 that you could put the slide under a microscope and then just capture it with a camera or just with a cell phone coupled to a camera, and then still make those predictions. So these models were quite robust to that kind of domain adaptation. And still I think that even today the slide digitization rate in the US remains at around 4%, and the standard of care is just looking at a glass light under a microscope. So it's very important to see how we can further democratize these models by just using the microscope, because most microscopes that pathologists use do have a camera attached to them. So can we somehow leverage that camera to just use a model that might be trained on a whole slide image, still work with the slide under a microscope?Eric Topol (06:12):Well, what you just said is actually a profound point that is only 4% of the slides are being reviewed digitally, and that means that we're still an old pathology era without the enlightenment of machine eyes. I mean these digital eyes that can be trained even without supervised learning as we'll get to see things that we'll never see. And to make, and I know we'll be recalling back in 2022, you and I wrote a Lancet piece about the work that you had done, which is very exciting with cardiac biopsies to detect whether a heart transplant was a rejection. This is a matter of life or death because you have to give more immunosuppression drugs if it's a rejection. But if you do that and it's not a rejection or you miss it, and there's lots of disagreement among pathologists, cardiac pathologists, regarding whether there's a transplant. So you had done some early work back then, and because much of what we're going to talk about, I think relates more to cancer, but it's across the board in pathology. Can you talk about the inner observer variability of pathologists when they look at regular slides?Faisal Mahmood (07:36):Yeah. So when I first started working in this field, my kind of thinking was that the slide digitization rate is very low. So how do we get people to embrace and adapt digital pathology and machine learning models that are trained on digital data if the data is not routinely digitized? So one of my kind of line of thinking was that if we focus on problems that are inherently so difficult that there isn't a good solution for them currently, and machine learning provides, or deep learning provides a tangible solution, people will be kind of forced to use these models. So along those lines, we started focusing on the cancers of unknown primary problem and the myocardial biopsy problem. So we know that the Cohen’s kappa or the intra-observer variability that also takes into account agreement by chance is around 0.22. So it's very, very low for endomyocardial biopsies. So that just means that there are a large number of patients who have a diagnosis that other pathologists might not agree with, and the downstream treatment regimen that's given is entirely based on that diagnosis. The same patient being diagnosed by a different cardiac pathologist could be receiving a very different regimen and could have a very, very different outcome.(09:14):So the goal for that study is published in Nature of Medicine in 2022, was to see if we could use deep learning to standardize that and have it act as an assistive tool for cardiac pathologists and whether they give more standardized responses when they're given a machine learning based response. So that's what we showed, and it was a pleasure to write that corresponding piece with you in the Lancet.Eric Topol (09:43):Yeah, no, I mean I think that was two years ago and so much has happened since then. So now I want to get into this. You've been on a tear every month publishing major papers and leading journals, and I want to just go back to March and we'll talk about April, May, and June. So back in March, you published two foundation models, UNI and CONCH, I believe, both of these and back-to-back papers in Nature Medicine. And so, maybe first if you could explain the foundation model, the principle, how that's different than the deep learning network in terms of transformers and also what these two different, these were mega models that you built, how they contributed to help advance the field.Faisal Mahmood (10:37):So a lot of the early work that we did relied on extracting features from a resonant trained on real world images. So by having these features extracted, we didn't need to train these models end to end and allowed us to train a lot of models and investigate a lot of different aspects. But those features that we used were still based on real world images. What foundation models led us do is they leveraged self supervised learning and large amounts of data that would be essentially unlabeled to extract rich feature representations from pathology images that can then be used for a variety of different downstream tasks. So we basically collected as much data as we could from the Brigham and MGH and some public sources while trying to keep it as diverse as possible. So the goal was to include infectious, inflammatory, neoplastic all everything across the pathology department while still being as diverse as possible, including normal tissue, everything.(11:52):And the hypothesis there, and that's been just recently confirmed that the hypothesis was that diversity would matter much more than the quantity of data. So if you have lots and lots of screening biopsies and you use all of them to train the foundation model, there isn't enough diversity there that it would begin to learn those fundamental feature representations that you would want it to learn. So we used all of this data and then trained the UNI model and then together with it was an image text model where it starts with UNI and then reinforces the feature representations using images and texts. And that sort of mimics how humans learn about pathology. So a new resident, new trainee learning pathology has a lot of knowledge of the world, but it's perhaps looking at a pathology image for the first time. But besides looking at the image, they're also being reinforced by all these language cues from, whether it's from text or from audio signals. So the hope there was that text would kind of reinforce that and generate better feature representation. So the two studies were made available together. They were published in Nature Medicine back in March, and with that we made both those models public. So at the time we obviously had no idea that they would generate so much interest in this field, downloaded 350,000 times on Hugging Face and used for all kinds of different applications that I would've never thought of. So that's been very exciting to see.Eric Topol (13:29):Can you give some examples of some of the things you wouldn't have thought of? Because it seems like you think of everything.Faisal Mahmood (13:35):Yeah, people have used it to when there was a challenge for detecting tuberculosis, I think in a very, very different kind of a dataset. It was from the Nightingale Foundation and they have large data sets. So that was very interesting to see. People have used it to create newer data sets that can then be used for training additional foundation models. It's being used to extract rich feature representations from pathology images, corresponding spatial transcriptomic data, trying to predict spatial transcriptomics directly from histology. And there's a number of other options.Eric Topol (14:27):Well, yeah, that was March. Before we get to April, you slipped in the spatial omics thing, which is a big deal that is ability to look at tissue, human tissue over time and space. I mean the spatial temporal, it will tell us so much whether an evolution of a cancer process or so many things. Can you just comment because this is one of the major parts of this new era of applying AI to biology?Faisal Mahmood (15:05):So I think there are a number of things we can do if we have spatial data spatially resolved omic data with histology images. So the first thing that comes to my mind as a computer scientist would be that can we train a joint foundation model where we would use the spatially resolved transcriptomics to further enforce the pathology signal as a ground truth in a contrastive manner, similar to what we do with text, and can we use that to extract even richer feature representation? So we're doing that. In fact, we made a data set of about a thousand pathology images with corresponding spatial transcriptomic information, both curated from public resources as well as some internal data publicly available so people could investigate that question further. We're entrusted in other aspects of this because there is some indication including a study from James Zou’s group at Stanford showing that we can predict histology, predict the spatial transcriptomic signal directly from histology. So there's early indications that we might also be able to do that in three dimensions. So yeah, it's definitely very interesting. More and more of that data is becoming available and how machine learning can sort of augment that is very exciting.Eric Topol (16:37):Yeah, I mean, most of the spatial omics has been a product of single cell sequencing, whether it's single nuclei and different omics, not just DNA, of course, RNA and even methylation, whatnot. So the fact that you could try to impute that from the histologies is pretty striking. Now, that was March and then in April you published to me an extraordinary paper about demographic bias and how generative AI, we're in the generative AI year now since as we discussed with foundation models, here again that gen AI could actually reduce biases and enhance fairness, which of course is so counterintuitive to everything that's been written to date. So maybe you can take us through how we can get a reduction in bias in pathology.Faisal Mahmood (17:34):Yeah, so in the study, the study was about, this had been investigated in other fields, but what we try to show is that a model trained on large, diverse, publicly available data. When that's applied internally and we stratify it based on demographic differences, race and so forth, we see these very clear disparities and biases. And we investigated a lot of different solutions that were out there to equalize the distribution of the data to balance the distribution using or sampling and some of these simple techniques. And none of them worked quite well. And then we observed that using foundation models or just having richer feature representations eliminates some of those biases. In parallel, there was another study from Google where they use generative AI to synthesize additional images from those underrepresented groups and then use those images to enhance the training signal. And then they also showed that you could reduce those biases.(18:49):So I think the common denominator there is that richer feature representations contribute to reduced biases. So the biases not because there is some inherent signal tied to these subgroups, but the bias is essentially there because the feature representat…

Full show notes at the publisher

Euan Ashley: Exercise may be the single most potent medical intervention ever known Jul 05, 2024

Show notes

Recently, a series of papers were published in Nature and Nature journals illuminating the physiologic effects of exercise from an NIH initiative called MoTrPAC. To understand the wealth of new findings, I spoke with Professor Euan Ashley, who, along with Matt Wheeler, heads up the bioinformatics center.Earlier this week, Stanford announced Euan Ashley will be the new Chair of the Department of Medicine. He has done groundbreaking work in human genomics, including rapid whole genome sequencing for critically ill patients and applying the technology for people with unknown diseases. A few years ago he published The Genome Odyssey book. As you’ll see from our conversation, he has also done extensive work on the science of exercise.Video snippet from our conversation. Full videos of all Ground Truths podcasts can be seen on YouTube here. The audios are also available on Apple and Spotify.Transcript with audio and external linksEric Topol (00:06):Well, hello, it's Eric Topol with Ground Truths, and I'm really delighted today to welcome my friend, Euan Ashley. He is the Roger and Joelle Burnell Chair of Genomics and Precision Health at Stanford. He's done pioneering work in genomics, but today we're going to talk about something very different, which he also is working in exercise. Exercise the cover of a Nature paper in May regarding this MoTrPAC, which we're going to talk about this big initiative to understand the benefits of exercise. But before I hand it over to Euan, and I just want to mention his description of the paper that he posted to summarize started with, “Exercise may be the single most potent medical intervention ever known.” So Euan welcome.Euan Ashley (01:01):Yeah, well, great. It's wonderful to be here, Eric, and so nice to see you.Eric Topol (01:06):Yeah. Well, we have a lot to talk about because exercise is a fascinating topic. And I guess maybe we'd start with the MoTrPAC, which is an interesting acronym that you all came up with. Maybe tell us a bit about that with the 800 rats and the 2,400 people and the 17,000 molecules, there’s a lot there.Euan Ashley (01:24):Right, right. Yeah. Well, first of all, of course, before you do any scientific study, especially with a large number of people in a consortium, you need a good acronym. So that was where we started with the idea was to focus on the molecular transducers of physical activity. As you pointed out there at the beginning, we really don’t have a more potent medical intervention, especially for prevention of disease. I mean, it’s just such a powerful thing that we have, and yet we don’t really understand how it works. And so, the MoTrPAC Consortium was designed to really work together, bring groups of people across the US together who all have some interest in exercise and some ability to measure molecules and really put together the world's largest study of exercise to try and start answering some of the questions about where the potency of this intervention come from.Eric Topol (02:20):So the first crop of papers, and there were several of them that came out all on the same day in Nature publications, was about the rats. The people part is incubating, but can you give us a skinny on, there was a lot there, but maybe you could just summarize what you thought were the main findings.Key MoTrPAC FindingsEuan Ashley (02:43):Yeah, of course, of course. And the MoTrPAC Consortium, I'll say first of all, yeah, large group is probably I think 36 principal investigators funded by the Common Fund. And so, it brings together large numbers of people, some of whom who spend most of their time thinking about let’s say animal exercise. Some have spent a lot of time thinking about humans in exercise and many of whom think about measuring technologies. And as you say, these first group of papers were focused on the rat study, but actually the study goes much more broadly than that. But of course, there are some advantages to the animal protocols. We can look at tissue and we'll talk about that in a moment. But the humans, of course, are where we're most interested in the end. And we do have tissues coming from humans blood and adipose tissue and skeletal muscle, but those are obviously the only organs we can really access.(03:31):So there's a rat study, which is this one we'll talk about, and that's aerobic exercise and training. There's human studies that include aerobic exercise, strengths studies as well. There's a study in kids, pediatric study and then also a study of people who are very fit because here we're focusing on the change from sedentary to fit. And so that gives us the key exercise signal. So this first crop of papers was really our first look, cross-tissue, cross multi-omics, so multiple different modalities of measurement. And I think, yeah, we were like about nine and a half thousand assays, 19 tissues, 25 different measurement platforms, and then four training points for these rats. So let's talk about the rats for a minute. What do they do? So they normally live at night. They're active at night. In this study, we reverse that so that we can actually do the studies during the day.(04:25):So we reverse their at night cycle and they do their treadmill exercise over the course of several weeks. They start with about 20 minutes, and they do more every day. There's a control group of rats that just get placed on the treadmill and then don't do any exercise. And so, this is a controlled study as well. And over the course of time, we work more, it's about eight weeks in total and then two days after each of those bouts of exercise. So it's not an acute study, we measure to see where we are. So we also have this time trajectory of exercise. So what did we find? I mean, I think the first thing I would say, we talked about just how potent exercise is. It's very, very clear from looking at all these tissues that when you exercise regularly, you are just a different person, or in this case a different rat.(05:15):Like literally every tissue is changed dramatically and some in quite surprising ways. So I give you a couple of the things that surprised me or that I thought were most interesting. The first thing was this question of how does exercise actually work? Because exercise is a stress. You go out and you pound the pavement or you're on the bike or whatever, and then your body recovers. And so, there's been this idea, it's referred to as hormesis, this idea that some of the benefit of exercise might come from this recurrent stress. So your body learns how to deal with stress. And so given that we were very interested that this heat shock response was so prominent across multiple tissues. So heat shock proteins are molecular chaperones and they take care of protein folding to make sure it's appropriately done and they prevent protein aggregation. And when proteins need degraded because they're damaged, the heat shock system jumps in.(06:10):So perhaps not surprising, but pretty interesting that the heat shock proteins were very prominent part of the stress response to exercise. And remember, this is not acute exercise, so these are benefits that are built up over time, so that was one. A surprising one to me, the adrenal gland. So we're used to thinking of adrenaline as an epinephrine, as a stress hormone, but actually we saw dramatic changes in the adrenal gland and we don't necessarily think too much. You think about the exercising muscles, you think about the heart, we think about the lungs, when we think about exercise, you don't necessarily think that you're changing your adrenal gland, but it was one of the most changed tissues. The immune system was a common upregulated system. We saw that. And in fact, some of the tissues in which the immune genes were most changed were somewhat surprising.(07:02):So the small intestine, for example, was a place where there was a highest enrichment of immune mediated pathways. And then some tissues changed pretty early, like the small intestine changed after just one or two weeks of training other tissues like the brown adipose tissue. It was more like seven or eight weeks of training before we saw the real changes in there. So just one or two little things that struck out, but I think this really the first molecular map of exercise. So we're looking across the whole system across multiple modalities of measurement across multiple tissues.Simulating StressEric Topol (07:34):So as far as understanding the benefits of exercise, does this tell us that it really does simulate stress that it's conditioning the body to deal with stress as reflected by the various points you just summarized?Euan Ashley (07:51):Yeah, I think that is exactly right. I mean, part of what we were trying to understand was in what way are you changed after you do exercise regularly? And I think if we think about things that are positive, then the ability to deal with stress at a cellular level, quite literally repair mechanisms seems to be a big part of it. The other aspect that was interesting is that when you're measuring this many analytes, you can also compare that with disease. And so, we understand that exercises is preventive benefit against disease. So in some cases, and this was work highlighted by my colleague Maléne Lindholm in the mitochondrial paper that came along with the main paper and she looked with a team across all mitochondrial changes across all of the tissues of the cell. So these are the workhorses of the individual cells that like the batteries inside the cells of the mitochondria.(08:54):And we saw big changes across, it's not surprisingly, but it's the energy source for cells, big changes across many tissues. But interestingly for two specific really important diseases, a liver disease in one case and type 2 diabetes on the other, it was very clear that the training upregulated a network that was exactly the opposite of that of the disease. And so, it really was intervening in a way that was very specifically opposite to the way we know disease mechanisms go. So it does seem like, I mean people talk about an exercise pill. I think this shows that that is just not going to be possible. There may be ways we could mimic some elements of exercise, but there's no pill. This is a multisystem, multi-tissue, multidimensional response to exercise.Eric Topol (09:44):Yeah, I think it's really important. That was one of the questions I was going to ask you is whether this would ever be simulated by a drug. And I think you already answered that, and the fact that it's so comprehensively sweeping across every organ and all these different signals, tens thousand plus signals across them, it's really striking. We never really understood the benefits of exercise and not that it's all resolved by any means. Some of the things that were interesting too was the sex specific findings. Maybe you want to comment about that because we don't spend enough time thinking about how sex does have a big effect on physiology.Sex-Specific FindingsEuan Ashley (10:24):Yeah, I mean that's a really good point and one that I think was really underlined for us at every corner, every turn of the analysis here. So really no matter which measurement modality, no matter which tissue, no matter which point of training, if we just asked these computer models to sort of separate the data according to the prominent signals without giving it a clue of what to do, the so-called unsupervised models, then sex basically came out every single time. So I think you say you're absolutely right that we so often overlook the difference. For years we've said, oh, it's too expensive to do animal studies in both sexes, so we'll just pick one. And males were picked more often. But there are plenty of studies that were just females, and I mean that clearly is wrong, and we are really, sometimes it appeared like we're almost dealing with two different species.(11:18):They were so different. But I think we can also learn from what those differences were. Interestingly, some of them were most profound in adipose tissue, so in fat, and that was the case both at rest, sedentary and amplified by exercise. So we saw big difference between females and males in relation to the kinds of signals that were prominent in the white adipose tissue. So this fat storage tissue, for example, in sedentary females, insulin signaling and the trigger to make fat and store fat was very prominent. But whereas in the males, even before any exercise, the fat signals were more related to metabolism, and we could have wild speculation about in evolutionary terms why that might be. Obviously, males and females have different biological many differences in their biology and obviously thinking about hormone systems and specifically pregnancy of course. And so, we could probably come up with some theories. In reality, all we know now are these observations were found and they're pretty interesting and they show us that we really always need to think separately about both sexes and look at both independently.Eric Topol (12:39):Well, and the other thing that you already pointed out, but I just want to underscore, you can't do this stuff in people. You can't just do fat biopsies and whatnot. So I mean, the fact that you can do this multi-omic, multi-organ type assessment is just really an extraordinary opportunity for learning. And while we're on the white fat story just briefly, we would rather have a lot more brown fat, but as we age, and I assume it's the same in rats, they don't have much as they get older brown fat. Does exercise help us get more brown fat or are we just stuck with the white adipose tissue?Brown vs White FatEuan Ashley (13:21):Yeah, well, it certainly allows us to have less of a white adipose tissue, and I think it's potential that our brown adipose tissue maybe more functional, and for those who are listening who are not familiar, I mean these really are different colors that relate to the actual color of the tissue, but the color is different because the brown adipose tissue contains lots of mitochondria and lipid droplets, and the brown adipose is there to help essentially generate heat. It has a very different function in a way, but even white adipose tissue that we think of as just being about storing energy, people think of fat as a very metabolically neutral or inert tissue, but in reality it's not. It's signaling. It's constantly, it's a tissue that's as alive as any other and not just a storage for excess energy, but exercise definitely appears to alter both in this sexually dimorphic way as we noted already and clearly both in a positive health way where I think the makeup of the brown tissue is different. The white tissue, there is less of it obviously with exercise, which is something that is well known, but not new here for the first time. But still important to have seen that even in the rats.Eric Topol (14:49):And there's even, we talked a moment go about drugs, but there are some molecules that are thought to be able to help convert white to brown fat that are understudy and we'll see if they get anywhere that's interesting. But also, you talked about aerobic exercise and with us both being cardiologists, and I know throughout my earlier part of my career, we only talked about aerobic exercise. There was no such thing as strength training, and we even discouraged that or we never talked about it. Now we know how important strength training is and not just strength and resistance training, but balance and posture and all these other things. I assume you can't study that in the rats.Euan Ashley (15:32):Well, it's not imp…

Full show notes at the publisher

Our TOPPODCAST Picks

Follow Us

Stay Connected

Related Podcasts

Links

Stay Connected