A Bayesian Assessment Of The Origins Of COVID-19 Using Spatiotemporal And Zoonotic Data

Transcript

Andrew Levin:                   Well, thanks very much to Professor John Taylor for inviting me to Hoover for the last few weeks, and thanks to Bob Hall for hosting this workshop today. Can everyone hear me okay? Okay, so it's coming close now to marking the fifth year since the outbreak of COVID. And we know that it was a terrible event all over the world. And what's remarkable, in some way, is that almost five years later, we still don't really have a clear idea of where it came from.

                                             And so that's what this paper is about and that's what this workshop will focus on. And let me just emphasize that these are just my own views. I literally, everything you see in this paper, I did myself. All the data analysis, all the charts, all the maps, everything. So it's my responsibility and I'm really eager with this workshop to get constructive feedback and suggestions, how to improve this work.

                                             I have no axe to grind whatsoever, no conflicts of interest whatsoever. Really just honestly trying to take a scientific approach here as carefully as possible to try to help answer the question, where did COVID come from? I'm just going to say that in 2021, the World Health Organization did a review of the origins of COVID. It's published a report on the WHO website. So it's only a little over a year after the outbreak started. And at the time I thought it'd be preliminary, but WHO has not come back since then to do further work on it.

                                             Experts at CDC in the United States and the CDC in Europe and elsewhere, to my knowledge, have also not issued reports on this. Most of the reports we've had have come from national security agencies and intelligence agencies and some congressional reviews, but it's remarkable that there just hasn't been more scientific research on this question given how important it is, and given how catastrophic it was. So if nothing else comes from this presentation, I hope it'll spur a fresh look at these questions. Seems appropriate, again, as we approach the fifth anniversary of the start of the pandemic.

                                             One of the things that was remarkable to me is I started working on this several years ago, and by the way, this is my fourth paper on the epidemiology of COVID. The first three have a thousand Google Scholar Citations. So, I'm proud that I have made some contributions before this. But as I was thinking about this question, I was surprised that there’s a lot of emphasis on genetic data and molecular analysis of the COVID virus that was called SARS-CoV-2. And relatively scarce analysis of the pattern, the spatial temporal pattern of initial cases, and remarkably little analysis of what I'm going to call the zoonotic data.

                                             Zoonosis is a term that's used to talk about a case where an infectious disease spreads from an animal source into the human population. So, okay. So just to start here, this is the latest data I think as of yesterday, our world and data, which was a dashboard that I followed a lot for the first couple years. These are estimates of the total mortality from COVID based on the excess deaths. There are other methods, but I think this one's a fairly reasonable one. And you see the central estimate worldwide total is now over 25 million deaths. The Coven’s interval is somewhere between about 18 and 35.

                                             Confirmed deaths is lower, but that's because in many cases, especially in developing countries, they don't necessarily identify the cause of death. There's no autopsies, there's scarcity of viral test equipment and all sorts of things. And so the confirmed deaths is way lower than what was actually the death toll. Again, a very terrible catastrophic pandemic.

                                             So we know pretty clearly, I think there's now a consensus that the pandemic started in Wuhan, China. Early on, there was some-

Bob Hall:                            I just raise one point, and that is, has anybody looked at years, lost years instead of deaths? I hear so many deaths are-

Andrew Levin:                   Absolutely. In fact, one of my papers, which I'm very proud of, it's called “Prevalence and Mortality of COVID in Long-Term Care Facilities.” And that was one of the questions that we got was a peer-reviewed paper, was a measure years of life lost. And the answer is that unfortunately, COVID was very dangerous for middle-aged people, not just for ill and elderly. And so the years of life lost, it was horrible. Fortunately, young people and younger middle-aged people weren't as affected by it as much. So compared to say, the Influenza pandemic a century ago, that one mostly hit people in their twenties and thirties. Of course, there was a time in life expectancy was a lot shorter than now. So actually the years of life lost from that pandemic are probably not so different than this one.

                                             But anyway, those are really great questions we can come back to in the Q&A. What I wanted to say is that early on there are questions about whether there have been early COVID cases elsewhere in Italy or Western Europe. And subsequent analysis pretty definitively ruled that out. The pandemic started in Wuhan.

                                             Wuhan's a historic city. It's a beautiful city. I haven't been there, but it's been amazing for me to look at it and read about it. It's at the confluence of two major rivers in China, the Yangtze, which is the largest river I think in all of Eastern Asia and the Han River, which is one of its main tributaries, me and Wuhan. And that's historically why Wuhan's, one of the top 10 largest cities in China, very historic. That's where the pandemic started. And so the question I think people asked right from the beginning is, why did it start there? What's special about Wuhan? And in some way that question's a kind of a guidepost for this presentation and for this paper. And you'll see it turns out that question's really, really important of taking an assistive approach to looking at the origins of COVID.

                                             Okay, within Wuhan, there's a market. You can see this in Chinese, but I'm going to switch to the English translation. It's called the Huanan, which actually means South China seafood wholesale market. And about a third of the initial cases, which means during December 2019, before the New Year, the Western New Year, I should say. So before January 1st, a large fraction of the initial cases that were identified were linked to this market. And that was a strong reason why many scientists looking at these early signs and indicators and patterns suspected that this was a zoonotic outbreak. That it must have come from an animal, probably a live wild animal that was sold at the market.

                                             Now, why did scientists think that? Well, part of the reason was because there were two prior pandemics over the previous 20 years. One of them was the original SARS, which happened in southern China further south. And it turned out to be linked to an animal called the palm civet. And look it up on Google if you want to look, they look a little bit like a cat. They live in trees, they're solitary animals. It turned out that it wasn't the wild palm civets, it was palm civets that were being sold in markets in Southern China as an exotic kind of cuisine. And apparently the infection spread from bats to these big groups of palm civets in the process of transportation and in the markets in southern China. That's what happened. And it's very well documented. I think no one really questions that. By 2008, it was already very clear that that was what had happened. And by 2015 it had been really nailed down. Okay, so palm civets. Another one. One more thing. Okay, go ahead.

Speaker 1:                          [inaudible 00:09:10] these zoonotic cases, can you trace it back to the source?

Andrew Levin:                   Good question. So first of all, it's absolutely definitive that it came from palm civets. And the reason we can say that definitively is because they were able to do DNA analysis of, say... There's a lot of discussion this in the paper, by the way, restaurant workers and things, but they were able to do DNA analysis that confirmed the genetic sequence in a palm civet was identical to the one in the person who was exposed to it, things like that. And the timing of things.

Speaker 1:                          My thing is, if it's typically the case that they can identify zoonotic source with great confidence and it raises an amazing perspective, is that the explanation for the Wuhan outbreak?

Andrew Levin:                   Absolutely. Very good question. And so I think especially now that we're five years later, okay, because in the case of the original SARS, 20 years ago now, it didn't take long. Within a year they already were pretty sure it was palm civets. By the way, raccoon dogs was another suspect, but it became pretty clear it was palm civets. And then a couple of years after that, they figured out that originally came from bats. And so just, I might as well introduce the terms now, bats are called the host reservoir. And the virus circulates among bats. A lot of them live in colonies, sleep in caves, and they come out at night. And so viruses circulate in the bat colonies and many times multiple species of bats living in the same cavern and different viruses circulating, and the bats have really amazing immune systems that are good at resisting to the viruses circulating, but they don't have pathological effects on the bats. Okay, host reservoir. Important term in zoonosis.

                                             The palm civet is what's called an intermediate species. An intermediary because they're getting the virus from bats, and probably it was through bat droppings and bat urine that was dropping on the... Imagine all these palm civets in their cages and containers and stuff, and the bat droppings are falling and they get the disease, the original SARS. And then the palm civets, you have to imagine in Southern China at that time, when you walked into a restaurant in Southern China, there would be a stack of cages selling live palm civets. Sometimes 5, 10, 15. Like we think of when you go to Maine or Massachusetts with lobsters and you walk into the restaurant and you're like, "Oh, I kind of like that one. What's his name? Well that's George. I think I'll have George." Right? And so that's the way the palm civets were being sold. And so it's not so surprising when we know all these details that was a very clear zoonotic outbreak.

                                             Okay, another one. MERS, M-E-R-S, Middle Eastern Respiratory Syndrome. That one was identified in 2014, very strongly linked to camels. And most of the people who got MERS, actually, because it's not so transmissible as SARS was. Most of the people who got MERS had direct contact with camels. Turned out MERS isn't really spread very much through airborne transmission. It was mostly spread through direct contact with sick infected camels. But again, they traced it back pretty definitively, that probably MERS started in bats somewhere 40, 50 years ago. And then the camels were kind of a longer term intermediary host, but the original source of MERS was bats. So there was lots of research during the 2010s looking at bat viruses and how they spread to animals and what are the risks to humans.

                                             Okay, so again, that's the background here. When it was identified that a third of the cases in Wuhan were linked directly either to vendors or people who were buying products at this market. Even a couple cases where people just were walking through it, were linked to the market. And so people suspected, oh, this looks just like the original SARS. It was part of the reason why it's called SARS-CoV-2 is the name of the virus, but also because it has a lot of genetic similarities to the original SARS.

                                             Now, before we go on, I just want you to notice two things here in this picture. One is the word seafood. This was not a wild mammal market. In fact, there are 676 vendors at this market. This is all in the WHO report. And only about eight or 10 of those vendors, so we're talking here about a very, very small fraction, eight or ten, we're selling live mammals. Most of the other vendors we're selling shrimp, fish, some processed meats, vegetables, you name it, frozen foods. The other thing to notice here is the word wholesale. And it's been amazing to just, absolutely amazing to me five years later that the word wholesale has been just practically ignored in all the discussion. Maybe because most of the people like me, don't speak Chinese, not familiar with Chinese, and everyone just calls it the Huanan market, but it was a wholesale market.

                                             Why does that matter? Well, we know five years later, we know that COVID was mostly spread by people congregating in a small area with limited air circulation. And so when I first heard about Huanan market, I was kind of imagining like a supermarket with lots of crowded people and helping each other. That was not the case here. The Huanan wholesale market, 676 vendors. Someone want to make a guess how many daily visitors there were to this market? Call it 700 vendors. How many daily visitors were there to the market?

Speaker 1:                          10,000?

Andrew Levin:                   Yeah, a little less than 10,000. Okay. We know that from social media stuff and other sort of GPS tracking and things, about 10,000. Okay so that means each vendor is making about 10 sales a day.

                                             And these are sales mostly to people running restaurants in that neighborhood who are coming to the wholesale market to buy a bucket of shrimp or a big load of vegetables or whatever they need for their restaurant. And they're not hanging around, of course, once in a while maybe, but mostly just coming in like, okay, the shrimp's already packed up and ready. So they just come in, pay for it and walk out.

                                             This doesn't fit very well to what we kind of intuitively think of as a hotspot. Hotspot places it's when we had a bunch of people in a small, like a choir rehearsal in Washington state in a room this size of people singing at the top of their voices, and half of them got COVID by the end of that choir rehearsal. This is not that. Not at all. Okay, by the way, I'm going to try to talk for 45 minutes and then they'll be plenty of time for discussion. So let me keep moving forward here.

                                             This is a map of the market. Again, a little bit ironic. I'm just puzzled sometimes at these things. By the way, I'm a student of John Taylor's, and so things that I got from him is strong focus on transparency, systematic policies and procedures. The World Health Organization has a responsibility to the world. Their report that was published in early 2021 did not have these charts. And the only reason we have them is because one of the people who was on the international panel is a very distinguished scientist. I want to say Netherlands, I hope that's right. Dr. Marion Koopmans, she gave a talk later in 2021 where this is coming from one of her slides. So evidently they made available materials to that panel that were never published for the general public. And it seems ironic to me, still to this day, I have not seen this on the WHO website.

                                             But the key point here, and I'm sorry Bob, it's small, and I would've used the laser, but I won't, okay, is in the upper right is the wild animal meat. You see that label. And this is December 15th. And this is the assessment of the WHO based on talking to interviews with vendors and with other documentation. Which of the stalls we're selling animal meat? And it's about eight. There's some little brown spots, but ignore those. Those are the environmental samples. It's the big squares that are filled in that are the wild animal stalls. So on the left, upper left is aquatic products. The lower left is meat, which means in this case, beef and pork. And then on the right lower is the poultry. Okay?

                                             So again, as we would expect, it's a wholesale seafood market. It's not a retail supermarket. And it's not a restaurant like in Southern China with SARS filled with wild animals. They're kind of confined to one corner. Now, this is a menu that was posted before COVID, but someone took a photograph of it.

                                             And so in January 2020, right when the pandemic was breaking out, this came online. The source here is JustTheNews, but you can find it elsewhere. The bottom, the little print there is the information about the vendor who was selling these products. But what's remarkable here, when I looked at it is, wow, there's a camel. I'm like, really? They were selling camels in Wuhan. How'd they even get them there? But with some little help from various sources, I did translations of all these items, and it turns out that they're offering camel meat and I think maybe a camel leg.

                                             There's other exotic species here, but the notable thing is that they're offering palm civets. Okay, but in fact, we know from other work that there were no actual palm civets being sold at the market. This is a menu where it's sort of like, it's a wholesale market, remember? So someone says, well, I'd like to get a palm civet. They're like, okay, we can get one for you. We'll get a hold of it for you. But there weren't any actually on sale at this market.

                                             Other species, snakes and birds and so on that we know now are not susceptible to COVID. They weren't quite sure about that at the time. There's an exotic animal called the pangolin. It looks, it's an anteater with scales. There was a lot of concern that might've been the source, but there were no pangolins sold at this market either. There's only 10 mammals sold at Huanan Market.

                                             Some of them were rolled out pretty quickly. The ones that were left, if I remember right here, it's in the paper. But I think a badger seemed very implausible because badgers are burrowing animals. So, out with the badgers, and they weren't selling enough of them. So again, it didn't look like what happened in SARS back in 2004.

                                             What's not on this menu at all is a raccoon dog. But it turned out, people noticed this and noted it right at the time. There was raccoon dogs being sold at Huanan Market. And let me just emphasize, not very many. Probably about one a day. So this is a photograph taken in 2014 at the same market, and probably even at the same wild mammal stall of a lone, looks kind of lonely, raccoon dog there.

                                             And you imagine, again, a vendor who's got one in the corner of their stall just in case someone comes in, wants to buy it. Or maybe there's a particular restaurant that uses raccoon dogs and they come in and get one every day, but one a day. It doesn't sound again, anything like the palm civets with SARS or really the camels with MERS. But this was the prime suspect. And there's been papers written in top science journals over the last few years about, oh, it must've been raccoon dogs. They're the only ones left. We've ruled out all the other suspects. If this was an audit, this is it. You're looking at it.

                                             This is a better picture of a raccoon dog. And again, if I only accomplish a few things in this talk today, one of them is to make you think about and appreciate raccoon dogs. They're beautiful animals, really lovely animals. They're wild animals, of course. It is a canine, okay? So that's why it has the word dog, but the face and some of the patterns of the fur looks a lot like a raccoon. And their habits are actually quite similar to raccoon. They're solitary creatures in the wild. They mostly like to live in forests. I'll come back to this in a little bit, but imagine that they're kind of like raccoons. They're scavengers. They're omnivorous, okay?

PART 1 OF 4 ENDS [00:23:04]

Andrew Levin:                   Imagine that they're kind of like racoons. They're scavengers. They're omnivorous. They are raised a lot in China, not for food. They're raised for fur because they have thick fur, like a fox, like a mink. And I'll show you the details in a little bit, but 15 million raccoon dogs were being raised in China in 2019, 15 million. So then we start to... Our eyes were like, "Oh, well that must be it then. It wasn't in southern China with the restaurants, it was the fur industry." So we need to look at that, all right? But again, we want to learn about, and I want to emphasize the name here at the bottom of the screen. Kaarina Kauhala is a Finish scientist who was really one of the world's experts on raccoon dogs, and unfortunately, she passed away in 2022. So I never had a chance to talk to her or meet her or even talk to her on Zoom, but she's really a remarkable person and helped the world understand more about the species.

                                             So zoonotic, that's what we're going to call hypothesis Z. Z for zoonotic. That's going to be one of the two hypothesis that we're going to look at here. It's a Bayesian approach. The other hypothesis, which we have to consider is the possibility that Covid came from a lab, from an accidental lab leak. And the reason that seems possible at least, is because there was a lot of genetic research going on in Wuhan on bat viruses during the 2010s, including some pretty complex experiments I'll talk about briefly, but you can read about it in the paper and I'm happy to talk about during Q and A if you want. But what I wanted to show you here, all of you know, corona viruses have that name, including Covid and SARS and MERS. They're all corona viruses. The word corona is for crown, like the beer, it's a crown, and the crown has spikes.

                                             And what you're looking at here on the left side of the screen, no laser, Bob. On the left side of the screen is a spike. That's what the actual spike looks like. And this is some scientific work published in 2021 that has beautiful pictures in it. It's a pretty complex spike. What I want you to notice here is what's marked as S1, S2 site. It's a little kind of juncture there. And in order for the spike to join to a host cell, that site has to be broken. It's called cleavage. It has to be cleaved. And in SARS, the original SARS, and in MERS, and in other SARS related viruses, that cleavage of the S1, S2 site happens outside the host cell. At the point when the virus is getting close to the host cell, the host cell's own proteins, they're called protease,actually do that cleavage.

                                             It's not very efficient. What's amazing, remarkable, extraordinary about the Covid virus is that the cleavage S1, S2 happens inside the host cell when the virus is being reproduced. It's like a factory. And the factory isn't just producing copies of the virus, it's actually doing the cleavage of the S1, S2 site. Well, how does that happen? Well, the answer is because there's a sequence in the RNA of the Covid virus, SARS-CoV-2 for nucleotide sequence that enables a protease in the host cell. The protease is called furin, F-U-R-I-N, to cleave that inside the host cell. And so that means that when the virus arrives at a new host cell, it's like ready to go. Called a furin cleavage site.

                                             Now, what you want to keep in mind, again, you may forget this later, but at least remember it now for the next few minutes, it's called a poly basic cleavage site. Poly basic is referring to the pH, so acidic and basic. Poly basic means that there's multiple amino acids in the sequence of four nucleotides. And that poly basic sequence is not found in any other SARS related virus, none. It's actually in some other corona viruses, but not in any of the SARS related ones. So where did it come from? How did it get there? Boy I mean, because that's critical for human infection. If that four nucleotides wasn't there, they've done it by the way, in the lab, you can take it out with the genetic editing, and then you can't infect human cells anymore. Where did it come from? And because it's only for nucleotides, there's some worry that maybe that's what was done in a lab actually to insert those four nucleotides.

                                             I think I have two more pictures here. I'll just skip over these. This is showing you the rest of the process and why the cleavage matters because it has to do with how the virus joins to the host membrane. Now, just to give you some sense here about this, the sizes here are very hard for humans to imagine. So I had to come up with some comparison to help us think about this. Human beings and many other mammals roughly speaking, we have 3 billion nucleotides in our genome, 3 billion. And so I was thinking, all right, well that's the distance from the earth to the moon is 400,000 kilometers. I'm going to think of that as the starting point when we think about how big is the human genome. A bacteria is about four and a half million. So it's roughly a thousandth the size of the human genome. A bacteria is a simple, one cell creature, four and a half million nucleotides. If you think of the human genome is earth to moon, then the size of the E. Coli bacteria is like the distance from here to Los Angeles, more or less 500 kilometers.

                                             Now a virus by that comparison is tiny, just tiny. The entire SARS-CoV-2 virus only has 30,000 nucleotides. That's a short walk, four kilometers. I was comparing it to the distance of the Lincoln Memorial to the US Capitol, if any of you have walked on the mall in Washington, D.C. But we could think of it as probably some distance from here to somewhere in Menlo Park, maybe, I'm not sure. Within the virus, the spike gene, which is the crucial part that makes it able to attach to host cells, is 4,000 nucleotides. That's like 500 meters. That's the distance from here to Green Library. The furin cleavage site, which is the critical element that makes Covid able to infect humans is four nucleotides. And so again, if we keep with the same scale, it's the length of a keyboard, just slightly bigger than my laptop, half a meter that's like this.

                                             So you see here, we start to think about, all right, well, editing the human genome is really, really complex. Editing a virus genome is not so hard. That's why it was already being done in the 2010s. Editing a single gene within the virus is plausible. Inserting four nucleotides gets to be not so bad. And so another thing you want to keep in mind here is this field of gene editing has been moving really quickly in the last 10, 15 years. Really quickly, I was just talking to an undergraduate student who was out at Berkeley a couple days ago, and it's amazing what they're doing. Undergraduates now are doing gene editing. Why is that possible? With these two scientists here, you can see on the screen, won the Nobel Prize in 2020, what's called CRISPR/Cas9. We'll have to talk over in Q and A. It's just too complicated. But that they developed in 2012.

                                             And so that editing method was coming into widespread use in 2014, 2015, 2016. Then you start to think, huh, so maybe the timing of Covid in 2019 is actually not so coincidental because it's exactly when someone, like a graduate student or a postdoc in a lab says, "Ah, let me try inserting these four nucleotides and see what happens, and I'll grow up in a cell culture." And that graduate student ends up getting infected and then spreads it to family members or other people, and pretty quickly it's out of the lab, a lab leak. And there's more discussion of this in the paper, so I don't want to get too bogged down here, but just to say, this might've been unimaginable 20, 30 years ago. Now it's reality and whatever happened, I'll just say this now, whatever happened with Covid, we should all be aware here that the risk that a 20-year-old or a 25-year-old in a lab could create something that's incredibly dangerous.

                                             We know what happened 10 days ago with an AK-47. This is far worse and it's not over. It's just getting started, these experiments with viruses. I think the whole world really needs to sit up and pay attention and think carefully about how are we going to make sure that this doesn't happen again or it doesn't happen at all, is a really important question. Here's the Wuhan Institute of Virology. This institute was the first institute to have a... It's called biosafety level 4 lab. Those are the ones you've seen when people were working with the Ebola virus. It looks like an astronaut working in a containment lab where there's negative air circulation and so on to make absolutely sure that the virus doesn't escape.

                                             What's amazing about the WIV is that a lot of the research on bat viruses and genetic editing of bat viruses was done in what's called a BSL-2 lab. And you should think of BSL-2, I guess the expression people use is it's a little bit like a dentist office. BSL-2, when you're working with something pretty routine and you're just wearing a face mask and maybe some gloves and that's about it. And so the concern here, part of the reason why there's been real concerns about the possibility that Covid came from a lab leak and not from cenosis is because of that. It's because of the combination of the furin cleavage site that's really unique among SARS viruses and the fact that a lot of this research was being done in labs with relatively low biosafety standards. So now let's get into the details here. I'm going to try to be brief. Let's see where we got 12:35. So I think we're okay. Sorry, Bob?

Bob Hall:                            As I understand it, the Wuhan Institute had labs up to 4, not just 2.

Andrew Levin:                   They did, sorry, I just said that. Oh, sorry. Okay. I didn't explain it very clearly. The head of the Wuhan lab in 2019 wrote a paper that's published where he said, "We're using the BSL-4 lab only to analyze Ebola and Marburg." And we have other papers published by scientists at WIV where it says it's part of the method. It's in the method section. It says, "This work was done in a BSL-2 setting." So there's no controversy here. It's well documented that they had a BSL-4 lab. It's important. I think the fact that WIV was, and I think still is, the only laboratory in China that has a BSL-4 tells you that they were being given a lot of resources and ability to develop a lot of expertise in viruses. The risk of a lab leak though is not the BSL-4. The risk of a lab leak is the work on genetic editing of viruses that was being done at much lower levels of biosafety. Is that clear?

Bob Hall:                            You haven't mentioned the book title, if you want get all the details.

Andrew Levin:                   Yeah, actually I did look at the book. Sorry. Well, we can talk about it afterwards. By the way, he wrote that book, sorry, we're talking about Michael Lewis's book. It's called The Premonition. It was written, I think around the end of 2021. So I think part of what I'm going to try to do now, we're almost five years. What else do we know? He did not go through the scientific literature to see which experiments were done of BSL-2, BSL-3, and BSL-4.

Bob Hall:                            Well, he covers... I don't know. You're in charge here.

Andrew Levin:                   Not in charge, I want to answer questions. Okay? But I'm just telling you that we know with absolute certainty that some of these experiments with gene editing of that viruses closely related to SARS were done in BSL-2 labs at WIV. That is just unarguable. And if he didn't mention it in his book, it's because he didn't know about it.

Bob Hall:                            The important facts of the book are exactly in that area. As I remember, it's a little more complicated.

Andrew Levin:                   There are complications for sure. I just don't want to get bogged down here too much, but there's important complications that maybe need to be incorporated. Let me try to go through this and maybe they can come back to it a little bit. Okay, two hypotheses, we call them A and Z. Pretty simple. A accident, Z zoonotic, both hypotheses. We're going to assume it's bat related corona virus. If you want to start talking about the possibility of a risk of an influence epidemic, it would get into other things. And to make this a practical Bayesian exercise, I thought that at least was a sensible restriction to both hypotheses focused on bat related corona viruses.

                                             But the difference between them is was it an accidental lab leak or was it a zoonotic spillover from a wild animal? Now, the nice thing about Bayesian analysis, many of you know this is, it's good for evaluating two competing hypotheses because you start with priors, which may be based on intuition, or it may be based on prior evidence that you have. And then you update those priors with specific sources of evidence and you end up with a posterior. And that's what we're going to do in the paper. That's what we'll do here in this workshop, is we're going to start with priors and we're going to update them, but we got to clearly identify the two hypotheses.

                                             The difference here from classic cystic, classic cystic takes a null hypothesis and it tests whether that null hypothesis can be rejected. And there's two problems in that. One, if you do reject, it's not clear what the alternative is because you're simply rejecting a null. So what is the alternative? And if you don't reject the null, then that doesn't necessarily mean that the null is true. It just means you may not have had enough evidence to reject it. Unfortunately, some of the work on the origins of Covid used classical cystic, and people have criticized it like, "That doesn't really make sense in this case." If we really want to take seriously two competing hypotheses, then Bayesian analysis is the way to go. And so that's what we're going to do here.

                                             All right. This is the only equation I'm going to talk about today. There's a few others in the paper, but really trying to make the paper and this presentation non-technical so that it's accessible to a broader audience. Again, I think that was the lesson I took away from Professor John Taylor is keep it simple, make it clear. Think about the different source of uncertainty. So rather than one complex Bayesian model with a hundred parameters in it, what we're going to do is break this question into smaller pieces. I'll call them bite-sized pieces. They're still pretty complex, but at least four smaller components where we can then systematically look at each of these. And then it turns out here again, a nice feature of Bayesian analysis is that the posterior odds is the prior multiplied by these components because each of the components is essentially an independent conditional question.

                                             One, what's the marginal likelihood that the pandemic would occur in China in the People's Republic of China, which I'll say PRC from now on? That's number one. Why did it happen in China? We know that it happened in China. Is that likely under each of the two hypotheses? Two, why did it happen in Wuhan? We know it did happen in Wuhan. That's the data we're using. And the marginal likelihood means under each hypothesis, what's the probability that we would observe the pandemic starting in Wuhan. Three, we have evidence from the WHO about the spatial and temporal pattern of the early Covid cases at the market. There's been some classical statistical analysis of that, but no Bayesian analysis as far as I know, looking at that data five years later. So it seems really important to do this, and maybe someone could do a better job than I have, but at least to frame this question is an important one to look at.

Bob Hall:                            The institute is quite a distance within Wuhan away from the market.

Andrew Levin:                   Okay, good question. Very good question. I'm sorry we didn't talk about the geography much yet. But let me just say briefly so you have it in mind. The BSL-4 laboratory that was looking at Ebola and Marburg is a long, long way from the downtown area of Wuhan. But hypothesis A is not really about a leak from the BSL-4 lab. I don't personally think that's plausible, but the WIB had BSL-2 and BSL-3 labs. I think they still do, near Wuhan University right across the river from the market. And one of the... precisely component form, Bob, I'm sorry, you just interrupted me exactly the wrong point, but we'll treat it as a question. Okay, let's look at the pattern of early cases that were not linked to the market.

                                             And again, there's been some classical analysis of this, really simple analysis, but no Bayesian analysis of it. And the pattern is actually really striking. I'll show it to you hopefully in a few more minutes, but just think about why does this matter? We're just focused on component four for a moment. If you thought that this was a hotspot at the market, one raccoon dog that had the SARS-Co virus passed it on to a vendor or some other people who are going to that stall. And that's the starting point for Covid, then what you'd expect, we know this from epidemiology and human diseases, is you start to see things spreading out from there. And there's some randomness, but that's kind of a center point.

                                             And because most of the people coming to the market because it was a wholesale market, are people from restaurants in the area and some people who are selling products to the market. So it might be a farmer or someone who's catching fish in the lake and bringing them to the market. We would expect to see, if it's really a zoonotic, we'd expect to see some pretty distant cases right away because those are people coming into the market for specific reasons. And we'd expect to see a lot of really close cases right in that neighborhood of where the restaurants are and people going to those restaurants. And we do, we see a lot of that.

Speaker 1:                          [inaudible 00:44:00] Is it possible, if it's a lab leak and the first person who catches it is asymptomatic and never identified, then it's going to look like... [inaudible 00:44:15]

Andrew Levin:                   This is what Bayesian-

Speaker 1:                          Yeah, but [inaudible 00:44:17].

Andrew Levin:                   Let me answer your question. So in a Bayesian method, it might be that component four is not helpful to us. We can't distinguish the two hypotheses, but what I'm trying to say to you is actually it turns out to be pretty important. And the reason is because the unlinked cases, those are the ones where, as far as the China CDC could tell from many interviews and looking at various kinds of information, and WHO verified all this is that about two thirds of the cases, there was no known link to the Huanan market. Now, if those people were also being indirectly, secondary, tertiary kind of spread from the market, then the pattern of the linked cases and the pattern of the unlinked cases, it would look roughly similar, right? It's the people in the neighborhood, someone gets it at the restaurant, then they spread it to their neighbor, or they're going to that restaurant, they live in that neighborhood.

                                             So this goes back to the very origins of epidemiology in the 1800s, I think John Snow. But...

Speaker 1:                          Okay, but-

Andrew Levin:                   Sorry, let me finish. But if it's an accidental lab leak, the first person who got it would be asymptomatic, but they have neighbors, some of whom are older, and there's going to start to be symptomatic cases on the east side of the river. And now we're back, I think I forgot who asked it before. Was it you Steve, or was it you Bob? The WIV BSL-4 lab is way south, but the WIV BSL-2 and 3 are close to the university. They're in neighborhoods. And so what we'd expect to see under A, hypothesis A, is unlinked cases that are happening on the east bank.

PART 2 OF 4 ENDS [00:46:04]

Andrew Levin:                   ... Unlinked cases, they're happening on the East Bank. Okay, so we'll take a look in a moment here, but I'm just telling you, I'm promising you actually, that this is probably not going to be useless information that there is, in principle, can be a spread between the pattern of cases under A, versus the pattern of cases under Z. And no one has looked at this before... I think a question you should all be asking is, why is this work being done by a macroeconomist who used to work at the Fed? And I think part of the answer to that is because I'm a Bayesian, I like Bayesian work, and part of it's I'm a curious person and part of it's because I was well-trained by John Taylor to think about situations where there's a lot of model uncertainty and various other factors here. But I've just been constantly frustrated why this work hasn't been done already by epidemiologists and people at CDC and WHO.

                                             But that's the fact, it hasn't. And so again, even if you're not persuaded by this paper, the extent to which there's still a lot of evidence about Covid that seems relevant for evaluating the two hypotheses, we should have hundreds of thousands of scientists around the world take a fresh look at all of this. Five years later, we still don't know the answer, and I think we deserve an answer. People who were affected by it and suffered from it, and the memories of the people who passed away from it, we should do our best to try to get an answer to this question.

                                             Okay, so that's the approach we're going to take. So let me go... I'm trying to finish in 12 more minutes... But I want to leave some time for discussion. So I think it's fine. You've got a lot of the basic ideas now, so this actually won't take us so long.

                                             I want to just say by the way that I think reasonable priors is one-to-one prior odds of one-to-one. Now we're called flat priors or diffuse priors. It means we're not going to take a strong stand. It's called letting the data speak. I think that's a reasonable approach here. But the reason we have to think about it a little bit, and there's a lot of discussion on it in the paper, is because there are people who do feel very strongly, "Oh, it must have been zoonotic." And they have priors, the WHO panel said it was extremely unlikely to have been a lab accident. Well, that would correspond to priors of a hundred-to-one in favor of Z. A hundred-to-one, okay?

                                             And on the other hand, people who've looked at the firm cleavage site and some other characteristics of the virus think, "Boy, this looks like an accident." And they think the odds are a hundred-to-one in favor of the accident.

                                             Well, that tells me again, all right, well let's split the difference. Some of you think it's a hundred-to-one, some of you think it's one-to-a-hundred, let's just stick with flat priors.

                                             But it's an important part of a Bayesian exercise to evaluate and assess before we look at these four components of the data, what's our priors going in?

                                             Okay, now, component number one, why did the pandemic start in PRC? It's a simple question and we can evaluate that question under each of the hypotheses. Remember, the phase factor is a ratio of marginal likelihoods. So the marginal likelihood of A means, the likelihood that it would've started in China, given that it was a lab accident. And the second one is, what's the likelihood that it would've started in PRC if it was a zoonotic event?

                                             And we can actually use data to evaluate each of those marginal likelihoods, and then we compare them with the ratio of those two is the Bayes factor. And that's kind of how this set of evidence, the answer to the question, "Why did it start in PRC?" Will be helpful to us, at least moderately helpful, in sorting through the questions about the origin.

                                             I got to be really quick here because it turns out this question's not the most important, but there is information about the global distribution of research on bat-related coronaviruses. A lot of it was happening in the United States, but the US has pretty strict biosafety standards, been in place for a long time. There can be accidents and mistakes and stuff, we're humans, but the US research is pretty well regulated. A lot of the other research was being done in China. And again, it's remarkable when you look at these papers written by top health officials, the head of the CDC in China in 2019 wrote a paper where he said, we have no regulatory framework for regulating BSL-2.

                                             So again, I don't know what Michael Lewis says about this in the book. I don't know if he saw that paper, but when the China CDC director acknowledges in 2019, we don't have any regulatory framework for ensuring biosafety in BSL-2 labs, or really for kind of determining what's going to be done in BSL-2 versus three versus four, tells us that the risk of a lab accident that's related to a bat-related coronavirus in China would be pretty substantial.

                                             I did some assessment in the paper, I'm going to just skip over this really quickly, but it turns out that if you use this approach, and I'm using GDP per capita as a way to assess the biosafety, there's other ways to do that, but probably there's a 50% chance, if we're in hypothesis A, and remember the hypothesis, A is simply that the pandemic started from a lab accident. The marginal likelihood from that is about 50%. We would not be surprised that if it was a lab accident, we wouldn't be surprised that the lab accident happened in China, because that's where all the research is being done, and that's where the biosafety standards weren't very high.

                                             Okay. The other marginal likelihood we have to look at is zoonotic. And here we can look at the distribution of bat species that have these kinds of coronaviruses. By the way, bats live everywhere in the inhabited world except for Iceland and Greenland, and Antarctica. Those are the only places in the entire world where there are no bats.

                                             So, if you start with that premise, then you would just say the probability of a outbreak in China is one seventh or one eighth of the world population lives there. We can probably narrow it down a little bit because the MERS-related viruses, those could have been the source of a pandemic. Those are not quite everywhere. They don't circulate apparently in North America, Canada or the United States. They don't circulate in Australia. They don't circulate in the southern part of South America. But, they don't circulate in all of China either. So this map basically tells you it's probably about 20% chance. If it was a zoonotic outbreak of a MERS-related virus in China, it's probably about 20%. Okay? That's the marginal likelihood.

Speaker 2:                          Do you know... Do you have any information on the proximity of the humans in these zones to the bats?

Andrew Levin:                   That's discussed in the paper, and there's a lot of discussion of that in the 2010s. So absolutely, I'm not doing it here, because then you've got to start to have models and people have done that. And ironically, again, this is in the paper, and I didn't put up any slides on it, in the discussion in the 2010s, where did people think the next pandemic was going to happen? Answer, either Southern Africa because that's where Ebola and some of the camels from MERS were in parts of sub-Saharan Africa, or a lot of people really worry what's going to happen in Brazil because there's a lot of bat species in Brazil and the viruses are circling among them. There wasn't a lot of papers written about, "Oh, the next one's going to be in China." There were some, but that wasn't really the prime suspect prior to the pandemic.

                                             One more picture here, and by the way, there's an organization called International Union of Conservation Nature, IUCN. They're the world's experts on where species live. And so part of my job working on this project was to get all their data and make maps in Stata, which you can do and put together all the species. But anyway, this is the distribution of bat species that have SARS-related viruses. Now only eastern hemisphere. Not so much in Africa, there's some, we don't know if that's just because of under-reporting maybe there hasn't been enough explanation of SARS-related viruses, but it doesn't look like so much in Africa. But almost all of Europe, India, Iran, Turkey, all of Southeast Asia, pretty much, Japan, and not all of China. So again, what's remarkable, I have the table in the paper, but the probability, looking at this map of an outbreak related to a SARS-related virus of a zoonotic outbreak is about 20%. I think it's 19.6% or something. Okay, so one in five.

                                             Okay, now we can do our Bayesian exercise. Remember, the Bayes factor is the ratio of the marginal likelihoods. So the numerator is the A one, that was 50%, and the denominator is about 20%. Okay, that means two and a half to one. Now two and a half to one odds is not really considered very interesting for Bayesians, okay? It's called anecdotal evidence. It's like, all right, well. It shifted the odds a little bit in the favor of A, a little bit from Z. But if you started with strong priors about Z, this wouldn't be enough to persuade you. So we should move on.

                                             Next question. Why did it happen in Wuhan? This becomes a much more straightforward question actually, because under hypothesis A, we can think about where research on bat-related coronaviruses, and especially gene editing and so on, where was that happening in China?

                                             And you already know the answer. It was happening almost entirely in Wuhan. There were certainly researchers in other places in China, but that was the hub. It's like Silicon Valley of bats. In fact, the head of the bat research at WIB was commonly referred to as the bat lady or the bat woman. And there were lots and lots of stories. And there's a video that was posted by the Chinese television in early December, 2019, about a hero scientist at the Wuhan CDC who was going out to lots of caves and collecting bat specimens. And it's an amazing video to watch, only four minutes. There's a link to it in my paper. In the video, he takes off his gloves and masks in the bat cave. And a lot of people watching that is like, "Well, that doesn't seem like very smart." But I would say, "Oh, were you just for the sake of making the video entertaining," right? Probably doesn't do that usually.

                                             Okay? But the main point is that if you ask the question, and again, it's discussed in the paper, but what's the marginal likelihood that it would be in Wuhan, conditional on a hypothesis A and conditional on it being in China? And the answer is probably 90%, maybe 99%, but certainly very, very high.

                                             Okay, now we got to go to Z. Okay? We're going to focus on raccoon dogs. The paper has some discussion about other wild mammals at the market. We're going to just take for granted here that it was raccoon dogs, and we got to look at wild raccoon dogs and farm raccoon dogs.

                                             Wild raccoon dogs we know from Kaarina Kauhala's work over many decades. They are monogamous. They mate for life. Here's the two, the male and the female together, and it's taken at night with a trap camera. This is a baby raccoon dog. The dad actually takes care of the kids while the mom goes out to hunt and scavenge. They're really a very nice family, actually was really fun to read about all of this, but-

Speaker 3:                          I surely hope this is not Z?

Andrew Levin:                   What's that?

Speaker 3:                          He's now made me very biased against Z by praising these dogs.

Andrew Levin:                   As I said-

Speaker 3:                          We're not going to destroy family life.

Andrew Levin:                   As I said before, if nothing good else comes out of this talk, at least now you know something about raccoon dogs, and we can appreciate them, and I'll show you a little bit more in a moment here. But again, the wild raccoon dogs, it doesn't really fit very well because they're solitary animals. All the evidence we have about raccoon dogs is they typically, a pair of adults takes a square kilometer of territory, and they don't really interact with other raccoon dogs. And so it's hard to imagine how a wild raccoon dog would get infected from a bat. And it's almost inconceivable to think about how wild raccoon dogs would pass it on to other wild raccoon dogs. It just doesn't make sense.

                                             But even beyond that, we know from the work of these scientists who have been tracking raccoon dogs over decades, that... By the way, this data actually goes back to 1923, I think the first book about the range of raccoon dogs in mainland China, 1923 book. And what you see here is raccoon dogs live almost everywhere in China except for Shandong, which is that peninsula sticking out. And the western part of China where there's very few trees, okay, raccoon dogs live almost everywhere else. So again, if you ask the question here, Wuhan is a city of 11 million people, China is 1.4 billion, okay? The probability of a wild raccoon dog being the zoonotic source just from that alone would be 1%, and probably smaller than 1%. Because as I said, it's just difficult to think about how wild raccoon dogs would be an intermediate source.

Speaker 1:                          Some of these raccoon dogs are farmed and-

Andrew Levin:                   We're getting there.

Speaker 1:                          Do we know it was a raccoon dog? Or is that just your standard for all mammalian species that might have been, [inaudible 01:00:43].

Andrew Levin:                   Sorry, I'm skipping over this. In the paper... So there were surveys of the market done before Covid broke out, where independent scientists were surveying which wild animals were being sold at the market. And not just mammals, actually, they were surveying birds and reptiles.

Speaker 1:                          Were dogs the only one that was there?

Andrew Levin:                   No, there were 10 wild mammals being sold at the Hunan market. Okay? But several of those can't get covid. It's been shown that they-

Speaker 1:                          [inaudible 01:01:19]. Raccoon dogs are the only [inaudible 01:01:21]-

Andrew Levin:                   It's the only plausible. It's the only plausible one.

Speaker 1:                          Possible source at this market?

Andrew Levin:                   Yes.

Speaker 1:                          Okay.

Andrew Levin:                   Yes.

Speaker 4:                          And that's accepted, is it?

Andrew Levin:                   Good question. I mean, I think there should be more discussion of this, but the papers that are published in the top journals on science and nature are pointing to the raccoon dog, and those are the people who are in favor of Z-

Speaker 4:                          Okay, I just want to understand.

Andrew Levin:                   I don't know what the people who are A, it's hard for them to think about, "Well, if it was a wild animal, which one would it be?" But the people who are supporting Z have been pointing to the raccoon dog.

Speaker 1:                          Okay.

Andrew Levin:                   All right. By the way, this is the research center in Germany that does work on these issues. And I thought it was pretty cool picture because it's an island, and the work actually was done, I think in a BSL-3 lab, but on this island. And so the premise here, I think was that it's going to be contained. And of course, very careful biosafety standards. Okay? This is a small experiment published in 2021, and I'm not going to go through all the details. It's in the paper, and we can talk it if you want. But the bottom line here is, this experiment proved pretty convincingly. They used autopsies and PSCR and antibodies and whatever, serology. Raccoon dogs only spread covid directly through their nose, not through airborne transmission. And one reason you can see that in the picture here is the orange and red are the ones that got it, either inoculated or it was gone by their neighbor in the next cage.

                                             And the green and blue are the ones that didn't get it. And the green and blue are only one and a half meters away from the red and orange. So, if this was humans in a small room for a day or two days or a week, everyone would be exposed to it and a lot of them would get it. In this case, there's pretty clearly no airborne transmission. So again, even for the farmed dogs, I think it starts to raise some questions here, how that would happen.

                                             But now here's a couple pictures of the farms. When I look at this, I don't really see, well, how would bats dropping, of feces and urine get these raccoon dogs infected? It's not completely obvious, but it could be a different kind of farm, or it could be that it got into the food and the food was deserted.

                                             So we can imagine, at least you said Steve, fur farms, it's certainly much more plausible than the wild raccoon dogs. But big part here, raccoon dogs are slaughtered at the farm. They're not transported. So again, very different from Palm Civets. Remember, the palm civets are sent to a wholesale center, and then they're sent to the resale center, and then they're sent to the restaurant and they're in the cages in the restaurant breathing on the customers. Okay? Raccoon dogs are slaughtered at the fur farm, and then the fur is sent to for shipment.

Speaker 1:                          Okay? So you've convinced us, I think, well-

Andrew Levin:                   This isn't even the end here because, look, the fur farms are all in northeast China, hundreds of kilometers away from Wuhan. And the only possible story I can think of, honestly, and again, I'd like other people to think about this and look at it, maybe I'm missing something, but the scenario would be... There is a green area there. You see it? Hunan province, which is next to Hubei, where Wuhan is, has 1% of the fur farms of raccoon dogs in China, about 1%, okay?

                                             Raccoon dogs routinely escape from these fur farms. It's not that uncommon that they escape. So we have to come up with a hypothesis where a raccoon dog got infected with SARS-CoV-II at a fur farm, and then escaped and was trapped by a hunter who then sold it to the Honan market, and then someone at the market got infected with it.

                                             And that seems like a stretch. So you kind of jumped to this conclusion already, Steve. To me, it seems very, very unlikely. Now we can do a Bayes factor. Remember, our question here is why did it start in Wuhan? And the numerator was a lab accident, and the conditionality was given the pandemic having an outbreak in China. What's the probability of a lab accident in Wuhan? And we think that's probably 90 or 99%, call it a hundred to one. And our denominator is the probability that a raccoon dog farmed or wild was the source in Wuhan. And there the odds are probably less than 1%, and the ratio of those two is about 10,000. So just getting to component two, okay? Remember component one, which was, why did it happen in China? Wasn't all that helpful to us. Component two, 10,000 to one is already very strong odds for a Bayesian, probably you'd call it decisive, but we're not quite done.

                                             I'm just going to jump ahead here, the last two components, excuse me, spatial temporal pattern. This is the question. We started talking about the unlinked cases, and let me just show you a couple pictures here. There are some distant cases, many of which were linked. They were probably hunters or farmers or fishing people that came into the market long distances to bring their supplies. But if we zoom in here into the core of Wuhan, the magenta is the case that are linked to the market. As we expected, really as we expected. Almost all the pink cases are on the west side of the Yangtze River, because remember, those are the neighborhoods, the restaurants, the people living in those neighborhoods around the restaurants. And a few people coming in bringing probably because the more rural areas are the ones where... Sorry, the shading here is the population densities.

                                             The magenta that's a little bit further away from the core center is probably people bringing in supplies to sell at the market. The green or the unlinked cases, and there's plenty of green ones around the neighborhood, around the market. And it's not surprising because you figure out, well, if it's a secondary or tertiary infection in the neighborhood around the market, that would be completely consistent with the Z. Okay. But what's notable here is a lot of the greens are on the other side of the river, close to the WIV BSL-II and BSL-III labs.

                                             One more fact I should mention, which is that the Wuhan CDC was doing bat virus research. They were collecting specimens, they were doing genetic analysis. They moved their lab in the fall of 2019, and the new location was one block from the market. What starts to seem very plausible to me, I worked at a government agency for a long time. When you move, things get disrupted. Usual standards kind of get relaxed a little bit, if confusing, okay? And so the possibility, again, that this wasn't totally coincidental, that somehow, I mean, it might've been a graduate student in her postdoc, installing doing things in a hurry, that then it's not coincidental why there was a hotspot.

PART 3 OF 4 ENDS [01:09:04]

Andrew Levin:                   It's not coincidental why... There was a hotspot at the market, no question about it. But that could have been a transmission from a source that was a lab accidental leak. And that's what the green cases look like here. Okay, the other thing we want to... I'll just skip over this one too. What about inside the market? And under Hypothesis Z, it's pretty simple, we start with a wild animal, a wild animal infects the vendor or some other people who are buying from that stall, and then it spreads outward from there. And because very few people outside the market would've been infected elsewhere, you'd expect that this thing is spreading outwards. So, the number of cases outside the market shouldn't matter very much. And under A, it's the opposite, the cases are coming from outside the market in. And this is straightforward from an analysis, completely straightforward.

                                             And once I had this data, it took a long time to tabulate all the vendors and the [inaudible 01:10:11] has all the information and what date they got it, and which stall. I had to make maps and all this drawing a little bit on some other work, but that was a lot of work. But once I had it, you just run probate with live variable, the lagged. Where in the market did the previous infections occur? Okay, and where were the wild animal stalls? Now, just look here. First of all, this is December 13th. First case of a vendor at the market was nowhere near the mammal stalls. It was actually someone selling shrimp in a different part of the market. Okay, well, that's all right. Maybe it's just coincidental, but here's a week later, still pretty much the same. If anything, it looks like maybe that shrimp seller is spreading the infection to other people.

                                             This is the point where there's already quite a lot of unlinked cases outside the market, by the way. And just not very many cases inside the market. Here's a week after that. Now we're starting to see a lot more cases in the market. And a few of those cases are starting to look like they're close to the wild mammal stall. Okay, so now imagine this, I think I'm done really... Pretty much done. Okay, if you just took the cumulative cases. So in effect, just take December 27th, you'd say, "Oh, well, this looks like it could have come from the wild mammals because there's a lot of cases in that kind of lower left corner of the market." But if you say, "No, but we got to look at the spatiotemporal pattern," which means space and time, then these other pictures are really important. We shouldn't be overlooking them because those don't fit very well to the Z story.

                                             So, the bottom line here, and we can do this it's... I have all the results, I haven't written them up yet. I'm happy to send you the paper in another week or so when it's ready. But we got four components of the Bayesian analysis, multiplying a prior. Component one is odds of two and a half to one in favor of A. Component two is probably 10,000 to one. Component three is around a hundred to one, and component four is probably around 50 to one. So, just do it in your head, 50 times a hundred is 5,000.

Speaker 5:                          This was way over our [inaudible 01:12:42].

Andrew Levin:                   Okay? Times 10,000. We're talking about 50 million times two and half is a hundred million. In a court you'd probably willing to convict someone if the odds that they were innocent was a hundred million to one. That's these DNA tests where the defendant's like, "Oh, but it could have been someone else with the same DNA pattern." And the expert at this witness stand is like, "Well, no, that's a probability of one in a hundred million." Then the jury would probably convict.

                                             As far as I can tell right now from having done this analysis, this actually looks pretty decisive in favor of it, but I don't want to jump to any conclusion. I just want to stop here to say, "Let's talk about this. Let's think about it." It's an important question.

Speaker 6:                          So, if it comes from the lab, why wasn't it planned?

Bob Hall:                            Yes, that.

Andrew Levin:                   I don't have information. Then that becomes a national security question, and I don't have the capability or the information to-

Speaker 1:                          [inaudible 01:13:50] You can just call it hypothesis L.

Andrew Levin:                   Yeah, I could... Oh, L for what?

Speaker 1:                          Lab.

Andrew Levin:                   Oh-

Speaker 1:                          You're not-

Andrew Levin:                   ... I personally, okay, this is just a personal opinion here. I think what happened was, because again, I know these people. I think what happened was that a graduate didn't postdoc in one of these labs doing gene editing and totally underestimating the riskiness of what they were doing in a situation where these methods are developing very quickly, and they've got a lot of samples of that-

Speaker 6:                          Do we need analysis to do that, to sort of distinguish between planned versus accidental?

Andrew Levin:                   No, because to do the plain one, you have to have military intelligence. How would you know? And again I-

Speaker 6:                          I share your view, it was probably an accident, but none of your analysis, as I understand it really needs to take a stand on it.

Andrew Levin:                   ... It's a good question. Why don't we think about that some more? I think because this is intended to be scientific, and I don't want to draw on national security intelligence or any of those sort of things. To me personally, the idea that this was intentionally planned seems remote. It just seems remote. So, not to say that someone else can't consider it or look into it, this is how I framed this particular exercise.

Speaker 6:                          In communicating, what we're showing it not traditionally as analysis, it's... There's a lab so there's also on the other side it's not a zoonotic that you have compared to this. A raccoon dog in this market as the-

Andrew Levin:                   No, and sorry, we had to rush a little bit, but in the paper, I go through each of the other wild mammals that were being sold at the market. And let me just be totally clear here. There were other live animals being sold in the market, including live shrimp and live snakes. But we can rule all those out because there was lots of animal studies that showed that snakes can't get COVID and birds can't get COVID and fish can't get COVID. So, we're down to 10 mammals. Of those 10, I think, if I remember right, five or six cannot get COVID, period. They just can't get it, they can't transmit it.

Speaker 6:                          Outside the market, a zoonotic could be a human infection outside the market that then goes into...

Andrew Levin:                   Okay, interesting. That would be a really big shift among the people who've been advocating that this looks zoonotic. The paper in Science magazine, I think it was 2022, was titled The Huanan Market was the epicenter of the COVID outbreak. Okay? So, the scientists who have been saying all the evidence points towards Z, that's what they've been pointing towards. Now, maybe this analysis will help people rethink and say... Well, maybe like you said, I haven't looked into it because this was the focal point, but it's still like if you use the SARS and MERS as analogies here, and we have some others zoonotic outbreaks and we know something about how they work, you'd have to think a little bit. Again, and why in Wuhan if it wasn't the market? Okay, what other animal would it be and how would there be enough of them around to both... They have to be an intermediate host, right? By the way, I didn't say this either, but the bats generally cannot get COVID.

Speaker 6:                          Yeah.

Andrew Levin:                   Oh, sorry. Please.

Speaker 7:                          Andy. First of all, apologies for not being there in person. I had some minor back surgery this morning, so I'm laid up for a couple of days. There's some additional spatial information that you didn't mention that might shed some light on whether it was China and Wuhan, which is the distribution of the spread outside China. In particular, a huge early concentration of Chinese people leaving after Chinese New Year, going back to their jobs in Northern Italy. Which became a very, very big hotspot very early. So, I'm wondering if you're thinking of using any of that additional information rather than just the gross probability distributions that you used.

Andrew Levin:                   Yeah, I think for comparing A and Z, once you get past early January, the virus is starting to spread really widely and it wasn't being traced yet. And so, there's lots and lots more cases in Wuhan during January, February before they put in place a lockdown. And there was... Bob and I were talking about this. The diamond princess cruise ship where one passenger came on board with COVID and a large fraction, if I remember right, I think there were 600 infections out of 3000 people on the cruise ship. And that happened in the end of January, early February. So, we know it was actually starting to spread pretty quickly at that point. I just don't see how we would use information about Diamond Princess to learn about A versus Z. And again, maybe I'm missing something. I'd be eager if you can think about a way to use it. I don't want lose efficacy.

Speaker 7:                          I agree. Distinguishing A and Z, but I'm just saying you started with... What was the A prior... You started with a broader question. Is there some A prior reason to believe that it might've been China, it might've been Wuhan, and used some gross data on population GDP distributions of things. And it just seems to me that...

Andrew Levin:                   I'm sorry, I may have just gone a little bit too quickly, so let me try it one more time. Okay. We're starting with these two hypotheses, A and Z, the hypothesis themselves, they don't say anything... The hypotheses themselves don't saying anything about where, and then we.

Speaker 7:                          I agree with that. All I'm saying is, I think you would strengthen the early part of your discussion about are we pretty sure it happened in Wuhan? Okay?

Andrew Levin:                   Oh, I see. Okay.

Speaker 7:                          That's all I'm saying. Instead of saying, "Is it Z?"

Andrew Levin:                   There was, as I said, some question early on about did it really start in Wuhan? And part of the reason for that confusion is because there were antibody tests done of blood samples from Northern Italy and a couple other places where in the first wave of antibody assays, it looked like people had been getting COVID for quite a while before that. And then it turned out it was because the assays were having false positives. So, I think most scientists, from my understanding now agree that it's not really controversial that the outbreak started in Wuhan. So, there isn't much discussion from the paper, but I can certainly add it, and I can add an appendix that describes some of this information.

Speaker 8:                          I wondered whether on the spatial question, whether it was... I think it's probably not possible, but to get some data on where the people who work in the Wuhan lab live, because somebody could live with somebody working in the lab, could live with someone who works in the Wuhan market, seafood market. I mean, if this had happened in the United States or somewhere else, like my country Australia, there would be an investigation into all the people who work there.

Andrew Levin:                   I completely agree. And there's been several international commissions that have called for more transparency from the Chinese authorities to release more information that they may know. And I guess I'll just say there's been resistance from the Chinese authorities to provide any more information that was not already in the WHO report in 2021. And so, I agree with you that knowing more details, the usual case data kind of information would be really helpful. It might decisively answer the question A versus Z without any of the stuff I was showing you. Okay, maybe we would’ve already known this four years ago.

                                             Now, again, here's just my hypothesis, having looked through all this for years now very carefully, is that if it was a graduate student or a postdoc, we'll call them a rogue researcher for the moment. Doing a really novel genetic experiment by inserting the four sequence furin cleavage site into a virus that was otherwise pretty infectious, but not to humans, okay? And they observed that. Would make sense of a lot of other things we know. First of all, the head of the VAT research at WIV is a very distinguished, respected scientist who'd been working on this for many years, and who was very careful, I think. Okay. And she has said publicly that she checked and she couldn't find any records of anyone in her lab doing any kind of research with anything that looked like the COVID virus. And I'm going to give her the benefit of the doubt that she was telling the truth.

                                             If that's true, then it probably means that this thing might've happened from a postdoc in the Wuhan CDC during the move, which happened in fall of 2019. Then it would all make sense because that person just dumps all of their Petri dishes down the sink. They haven't really discussed their research with anyone else. And even the Chinese authorities can't quite figure out how did this happen? So, what you're saying about the case study that would be easier if we knew it was within a specific lab at WIV, say the BSL-4 lab, but if it could be one of a hundred or 300 different researchers doing experiments with genetic sequences of bat viruses, and it gets to be a harder problem.

                                             But again, the evidence now, I would say, at least from my interpretation of it, as of today, what did we say, a hundred million to one in favor of A? Okay, means that we should probably be going back again to say, "Well, how might this have happened?" And even more important than trying to figure out what happened then, how do we make sure it doesn't happen again?

Speaker 8:                          So, in December, your first date. I can't remember what your first date was, about the seventh of December, I guess. In December you know there was one person in the market with COVID working in the seafood area?

Andrew Levin:                   She was actually infected, sorry, December 11th.

Speaker 8:                          11th?

Andrew Levin:                   10/11, okay? She was infected. But remember, she's the first identified case that means that she was sick enough, she went to the doctor, ended up in the hospital. There might well have been earlier cases that were either asymptomatic or mildly symptomatic that the Chinese authorities wouldn't have any way to tell because that person might not have gone to a hospital or clinic.

Speaker 8:                          So, no one from the lab was identified?

Andrew Levin:                   No one was identified. That doesn't mean, I'm not sure that means much to us because we don't know whether the Chinese authorities would reveal that information if it was the case. And to the extent to which there have been any interviews of spouses or children or other people living with them, none of that information has been revealed. And as I say, international commissions have been urging the Chinese to be more transparent. And so far it hasn't worked. But maybe you could write a letter to the [inaudible 01:25:58].

Speaker 8:                          No, thank you. Okay.

Speaker 9:                          So, I just want to spell out why you think this is important. I'm sure you know this, but I think it's worth saying. That the importance of this question for the future is, if you really wanted to prevent this in the future, you might either just eliminate those beautiful dogs. Raccoon dogs. Or you might make sure you have very rigorous procedures in all laboratories. And so you need to know which it is in order to... But in a sense, wouldn't you want to have rigorous procedures in labs anyhow? So, does it really have any practical significance if you're [inaudible 01:26:48]?

Andrew Levin:                   I think it depends on how you think of how humans work. Because now we're back on my training with the Fed. When something bad happens is usually when people start making changes, okay? Trying to get people to make really significant, sometimes difficult changes when nothing bad has happened tends to be difficult. If you have children, teenage children, you kind of know what I'm talking about here. And so, I think from that point of view, probably it still could be important. But I would also say that in Britain, and I suppose that might be true in Australia, there's an inquest when someone dies and they feel like, "Okay, it's important to figure out what was the cause of death." And for that to become-

Speaker 8:                          Can be. Depends on what caused the...

Andrew Levin:                   Well, yeah, I suppose if it's a person who's 99 years old, maybe there's no inquest. But in cases where there's a real question, okay, and what I'm saying here is the inquest for the cause of COVID, which as we've looked at the beginning, probably 25 million people, maybe more have died from it. And we're not sure, then there needs to be an inquest. And the inquest can't just say, "Oh, we're just not sure." We're shrugged our shoulders and walk away. Part of my purpose today was to say there's a lot of data that we haven't really looked at yet and there are methods that haven't been applied to that data. And I think that the people who suffered from this pandemic deserve...

                                             I guess some scientists would call it a moral imperative or an ethical imperative to try to do the very best we can to give an answer to this question. Even if, in fact, our precautions about bio lab safety maybe shouldn't hinge on it. In fact, I think they probably do. But I still think we owe it to the public to try to give the best answer we can to this question.

Speaker 7:                          Okay. I would add a friendly amendment to your observation, Andy. Which is when bad things happen, people make changes, but then they depreciate over time. People get lax again over time.

Andrew Levin:                   Okay, but just today, okay, I opened my phone, I get the Google News, a new article published in Nature Today. Look it up it's about... There's a new method for gene editing that's like even better and faster and simpler. I forgot what the nickname was for it in Nature Today. And so this problem is going to get worse, not better.

Speaker 7:                          I agree.

Andrew Levin:                   And so the point at which high school students can insert a furin cleavage site for their science project. Okay, so then what would I think about, what we probably need to do is probably the software. The CRISPR software, which the United States actually produces a lot of it. Should probably have built-in protections that if it looks like... Like we have for computer viruses. Right? And so it scans when the high school student says, "I want to insert the four nucleotide sequence here." CRISPR says, "No, I'm not allowing you to do that. It's too dangerous." And not only that, but it sends an alert to the CDC like, "There's a high school student in Nebraska who's trying to insert a..." And maybe they look at like, "Oh, it's all fine," or maybe they go visit that person's house like the FBI does sometimes, and you know what I'm saying?

                                             So, we should probably be thinking about this very hard right away, regardless of the answer of A versus Z, because it's a looming risk. Well, look, I appreciate any comments, any suggestions, any criticism. This is totally new work. I'm really honestly trying to be as careful as I can. So, I appreciate your comments and questions in order to talk to you more. Thank you.

ENDS [01:31:12]