About that claim that “SARS-CoV-2 is not a natural zoonosis but instead is laboratory derived”

A couple people pointed me to this article, “A Bayesian analysis concludes beyond a reasonable doubt that SARS-CoV-2 is not a natural zoonosis but instead is laboratory derived.”

It is hard for me to assess this document, as the key issues involve the biology of the virus, and I don’t know anything about genetics. There are also some claims made about bats the geography of China, and I don’t know anything about that either. I do not think anyone should take the Bayesian analysis in the paper seriously—all those 1.2% and 98.8% and 0.2% and 99.8% numbers on page 6 seem completely arbitrary. In assessing the claims in this paper, I recommend going straight to the biology and epidemiology claims therein, and ignoring the supposed Bayesian analysis. This does not mean that the conclusions of this paper are wrong (or that they’re right), just that the biological arguments need to be considered on their own terms, setting aside the Bayesian distractions.

tl;dr. I’m not the one to evaluate this one. I can tell you the Bayesian stuff makes no sense, but I have not tried to assess the biological and epidemiological claims.

97 thoughts on “About that claim that “SARS-CoV-2 is not a natural zoonosis but instead is laboratory derived”

  1. My son, who actually studies viral evolution, says the Bayesian prior should very heavily favor zoonosis, because there are very many animal–> human transmissions every year. Without any formal argument, I tend to suspect a lab accident, just from many years of working in a variety of labs.

    • >>the Bayesian prior should very heavily favor zoonosis, because there are very many animal–> human transmissions every year.

      That makes sense. I think this event looks more unusual to us than it really is, because we have only really had the tools to detect viruses pretty recently.

      The four human “common cold” coronaviruses must have jumped to humans at some point, and it’s been suggested that the “Russian flu” pandemic in the 1890s might actually have been one of those events.

      Also, a COVID-like event 200 years ago might not have amounted to very much, because the population was far younger (thus far smaller at-risk population) and the baseline rate of pneumonia death, pre-antibiotics, was a lot higher.

      • “the Bayesian prior should very heavily favor zoonosis, because there are very many animal–> human transmissions every year”

        This would be true if you are talking about a virus like this emerging in no particular place in the world. And along the same lines, no one would be saying this came from a lab if it emerged…somewhere without a lab….or near a lab that didn’t study viruses of that type.

        If you structure the question properly – what is the probability of a virus with pandemic potential of the type studied by WIV emerging naturally in Wuhan – I think you’ll come to the exact opposite conclusion. It is exceedingly unlikely for a virus like this to emerge naturally in Wuhan. If it weren’t, we’d have pandemics all the time. There are a lot of cities, and from what I understand, Wuhan characteristics make it less likely than average for a virus like this to emerge naturally there.

        On the flipside, experts have estimated the risk of a laboratory-acquired infection in U.S. BSL4 labs of 0.3% per year. Diplomatic cables expressed concerns about the safety of WIV, so I think it is totally reasonable to assume the risk at WIV was higher.

        Needless to say, whatever the level of LAI risk at WIV, it is at least an order of magnitude higher than the risk of natural emergence in Wuhan. The conclusion should be an intuitively obvious one – if a virus of the type studied by a high security lab emerges in proximity to the lab, it is far more likely to have come from the lab than nature. If that were not true, we’d have dramatically more pandemics than we do.

        • Using likelihoods, consider Hypothesis L: escaped from Wuhan lab versus Hypothesis Z: zoonotic, and the Evidence E: the virus first appeared in Wuhan.
          The likelihood Pr(E|L) would be close to 1, that is if it escaped from the lab in Wuhan then it is extremely likely to first appear in the city of Wuhan.
          The likelihood Pr(E|Z) would be close to zero because there are so many different places where human/animal contact occurs.
          Hence the likelihood ratio Pr(E|L)/Pr(E|Z) is very large, probably more than 1000.
          Posterior odds = prior odds * LR
          Thus for one to favor Z over L would require prior odds in favour of Z of more than 1000:1.
          And that’s being pretty generous to Z.
          Likelihood doesn’t lie (!lieklihood)

    • Zoonoses are in fact incredibly rare given the huge number of human animal interactions that take place. A virus can’t merely “jump species” on casual contact. It takes a deep, prolonged exposure, the virus needs many generations to adapt and evolve, all while our immune system is battling it. To acheive highly efficient human to human transmission is rarer still.

      Lab accidents, on the other hand, are commonplace. SARS has escaped from labs on seven occasions. That is to say a ratio of at least 7:1 lab escapes to zoonosis in that particular instance.

      • “Lab accidents, on the other hand, are commonplace. SARS has escaped from labs on seven occasions. That is to say a ratio of at least 7:1 lab escapes to zoonosis in that particular instance.”

        But of course, they would not have been studying SARS in the lab if the original zoonosis event had not taken place.

        Assuming you’re even correct …

        • “They would not have been studying SARS in the lab if the original zoonosis event had not taken place.”

          The problem with the Zoonosis theory is simply that they have not found an intermediate host since the virus first emerged in China. That is officially more than an year now. WHO ‘investigations’ say that the Chinese tested 64000 animals but could not find the virus. Intermediate host was found for SARS1 in 10 months or so, they cannot find it because it doesn’t exist.

          Nobody has any problem with them studying SARS virus, everyone should be worried about them fiddling with live viruses in a BSL2 level lab. Even inactivated vaccines are produced in a BSL3 lab, this is industrial scale. But that ‘Bat lady’ even published papers in Nature journal saying, ‘Oh we did live virus study in BSL2 lab’, that is wearing a glove and lab coat. That is where the problem is.
          They simply must not be doing this. The whole argument for their research is we should be prepared for the new xt pandemic, clearly, after spending millions of dollars we were definitely not prepared.

  2. Three topics of interest:

    1. The biology arguments in favor a lab origin.
    2. The decision to glue on both a contrived Bayesian analysis with a bunch of made-up numbers, and a section on the philosophy of virus control, both of which serve to dilute the technical argument.
    3.The perceived need to somehow quantify the argument with numbers so that is seems – to some I guess – to carry more weight.

    • I think there’s a difference between laboratory derived (ie. that it was some how manufactured or altered on purpose) compared to laboratory *escaped* (ie. it was a naturally occurring virus being studied at a lab and whoops and accident happened) are two separate issues. I think it’s pretty dumb to think someone engineered this virus, but it wouldn’t entirely surprise me to find out that it was a naturally occurring varying that someone was studying, and accidentally infected some people who began to spread it. I don’t think we’ll ever know for certain. In many ways it doesn’t really matter.

      • I don’t necessarily disagree with this, but considering that we have no real evidence that it isn’t naturally derived, ultimately it comes down to your prior probabilities of a “natural” zoonotic transfer versus an accidental lab release of a virus. People can have justifiably different priors here, but I think that, given we know that there’s animal to human virus transmission all the time, and the number of confirmed lab escapes is relatively low, most people would IMO correctly judge the probability of a zoonotic origin to be far higher. But I agree that the answer is fundamentally unknowable at this point, and probably largely inconsequental, even if it’s a lab release, it doesn’t tell us anything we don’t already know about biosecurity.

      • Yeah, but I wouldn’t really expect a lab to be a particularly likely route for animal-to-human transmission, in a part of the world where the wildlife trade flourishes.

        If we were talking about a region with a much higher researcher-to-total-population ratio and less wildlife trade (like the Arctic or something), maybe…

        • Or, I guess phrased a little differently… would it make sense, given the virus having already arisen by natural mutation, for it to *first* encounter humans in a lab, given that this is a region with very dense human population and lots of human-wildlife interaction?

        • This part of China (Wuhan) does in fact gave a very high researcher to total population ratio. In fact if you looked at the researchers-doing-genetic-manipulation-of-bat-coronaviruses to total population ratio, it is the highest in the world. There are several labs forming part of the Wuhan Institute of Virology. One is the much discussed high biosafety BSL-4), but there are also several BSL2/3 laboratories, a university and a CDC also in the city, all working with bat coronaviruses.

      • I listened to an NPR show early on, probably April 2020, that interviewed bat biologists and bat virus people. They were all skeptical of the various lab theories. People studying infectious viruses are berry careful. The gear they wear to go into bat caves to sample viruses in China is extensive. But at the same time thr biologists are in the cave with all their protective gear, there would be tourists touring the cave with no gear on at all. Ditto for animal markets etc. and given the source of SARS-1, MERS, etc, there really is no need for some special unnatural event to cause SARS-2. “We have no need of that hypothesis”, to paraphrase Laplace.

        Unfortunately, some politicians etc do have a political need for that hypothesis, which IMHO is why it exists.

        • SARScescapes from one of these labs every couple years:

          SARS has not re-emerged naturally, but there have been six escapes from virology labs: one each in Singapore and Taiwan, and four separate escapes at the same laboratory in Beijing.

          https://nationalpost.com/news/a-brief-terrifying-history-of-viruses-escaping-from-labs-70s-chinese-pandemic-was-a-lab-mistake

          And those are just the ones we know about. It really is something that happens all the time. Humans are idiots for building BSL-4 labs next to major population centers. Put them in the middle of nowhere with a support village and quarantine before going back to the rest of the world.

          You will also get fewer careerists working on these things since it will be like a military deployment.

        • The SARS thing is not really comparable, though, as that is a virus already known from a prior (natural) outbreak, which is why it was in the lab.

          SARS-COV-2/COVID being a lab escape would imply that the lab was *the first place humans encountered the virus*; that’s what I’m saying seems fairly counter-Occamian – if the virus exists out there, in a population-dense region like Hubei Province, shouldn’t somebody run into it pretty quickly?

          It’s not the occurrence of lab escapes itself that is unlikely, really; it’s that being the “first” source.

        • SARS2 is almost surely derived from SARS or the same common ancestor. It is exactly comparable.

          Also, do some simulations with super spreaders, where ~10% of cases are responsible for ~80% of new infections. The majority of the time a new strain dies out before growing big enough to take hold.

          I get people want to pretend no one is doing bioweapon research but it really does not require much specialization as long as you are willing to hurt people and cover it up to achieve your goals.

        • >>SARS2 is almost surely derived from SARS or the same common ancestor. It is exactly comparable.

          Genetic similarity isn’t relevant here. SARS was in the lab since it was already known, so its escape only requires carelessness.

          A *novel* virus being a lab escape requires the scientists to find it *before* someone gets randomly infected with it (after which it’s no longer new). That seems much less likely.

          >>The majority of the time a new strain dies out before growing big enough to take hold.

          True, but I don’t see the relevance. (If anything, I’d think this argues against a lab escape, which would probably be just one or maybe two people exposed – vs. say an agricultural or wildlife market setting where many people could be exposed to the same animal.)

          >>I get people want to pretend no one is doing bioweapon research

          No way this is an intentional bioweapon, its traits are all wrong (too uncontrollable and not very dangerous to military, ie healthy young adults). I thought you were arguing escape from an academic/health lab studying viruses…

        • A *novel* virus being a lab escape requires the scientists to find it *before* someone gets randomly infected with it (after which it’s no longer new). That seems much less likely.

          What do you think they do in these labs? Just passaging the virus leads to selection of certain mutants depending on the cell type:

          Re-sequencing of the first 501Y.V2 variant after outgrowth revealed no changes in the RBD or NTD but a deletion in the furin cleavage site (Table S3) commonly observed after in vitro culture in Vero E6 cells [19, 20]

          https://www.krisp.org.za/publications.php?pubid=316

        • Here is an example of what is going on in these labs:

          Here we examine the disease potential of a SARS-like virus, SHC014-CoV, which is currently circulating in Chinese horseshoe bat populations. Using the SARS-CoV reverse genetics system, we generated and characterized a chimeric virus expressing the spike of bat coronavirus SHC014 in a mouse-adapted SARS-CoV backbone. The results indicate that group 2b viruses encoding the SHC014 spike in a wild-type backbone can efficiently use multiple orthologs of the SARS receptor human angiotensin converting enzyme II (ACE2), replicate efficiently in primary human airway cells and achieve in vitro titers equivalent to epidemic strains of SARS-CoV. Additionally, in vivo experiments demonstrate replication of the chimeric virus in mouse lung with notable pathogenesis.

          https://pubmed.ncbi.nlm.nih.gov/26552008/

        • “I get people want to pretend no one is doing bioweapon research”

          As usual, you argue from a false premise.

          That’s not the argument. And I’m guessing you know it.

          SARS and MERS are zoonotic in origin. There’s really no reason to suggest that SARS-CoV-2 isn’t, either – other than China bashing for the sake of bashing China.

        • Do not collect evidence to do a proper investigation, because it may be politically inconvenient for someone.*

          This is a stats blog, people here are supposed to like data.

          So many new anti-science posters showed up here last year, always wanting to avoid collecting evidence about this or that.

          * It is just as likely to have come from the US or a joint US-China project as far as I’m concerned.

      • The “lab escaped” theory is just as unlikely as the “lab derived” theory to me. Basically we are comparing the chance of virus infecting humans in two scenarios:

        1) a team of maybe 20 people spending limited time collecting the virus from bats and then handling it in a BSL4 lab.
        2) a billion people in south China and south east Asia, and their countless livestock and pets living in close proximity with bats on a daily basis.

    • It is ridiculously simple to make a virus more infectious or deadly in a given species. It took a couple months to get a version of sars2 that made mice sick.

      First study I just found was published in early May and there were probably half a dozen different times this was done in 2020:
      https://pubmed.ncbi.nlm.nih.gov/32511406/

      It is honestly something we probably shouldnt talk about in detail. What they do in that paper isn’t even the easiest way.

        • Any national government in the world can do it if evil enough. Lots of lesser organizations too.

          You can also probably go 90% of the way using human cell lines.

        • I don’t think anyone is claiming that it’s impossible though, just that there’s zero legitimate evidence that it happened in this case.

          The only thing that seems to hint at labs is the location — and that area is also known to be a hotspot for natural bat coronaviruses, so…

          Zoonoses happen all the time (probably more often than we know – IIRC some of the common-cold coronaviruses were only discovered in the 21st century, stuff that’s mild enough might never get noticed).

        • I don’t think anyone is claiming that it’s impossible though, just that there’s zero legitimate evidence that it happened in this case.

          This idea was being actively suppressed. Recently it seems to have become non-verboten for some reason.

          Zero “legitimate evidence” is obviously false. All the available evidence is wholly consistent with a lab source, and escape from BSL-4 labs happens every couple years (at least).

          Do you also think there is no “legitimate evidence” for correcting vitamin C deficiencies in someone with a positive covid test?

        • >>Zero “legitimate evidence” is obviously false. All the available evidence is wholly consistent with a lab source

          But also consistent with natural zoonosis

          >>and escape from BSL-4 labs happens every couple years (at least).

          Not of *previously unknown viruses*, that’s never happened AFAIK.

          The unlikely thing is not so much the escape itself, but that a lab is the *first human-virus encounter*. Generally the virus is in the lab because people already discovered it (due to people getting sick with it).

          >>Do you also think there is no “legitimate evidence” for correcting vitamin C deficiencies in someone with a positive covid test?

          What do vitamin deficiencies have to do with the origin of a virus?

          (Obviously if a vitamin deficiency exists it should be corrected. But widespread/preventative use of vitamins – especially C – is going to attract a lot of skepticism due to past wrong claims of wide protective value.)

        • But also consistent with natural zoonosis

          So then there is “zero legitimate evidence” it was a zoonosis as well, according to your use of that phrase.

          Yet one explanation has been discouraged and the other supported.

        • If the evidence does not distinguish* between two explanations, why go for the much more sensational and less likely one?

          Zoonoses happen *all the time*, why even look for another explanation?

          (*If the argument is for a lab escape of an essentially natural virus, I don’t think there really CAN be any solid evidence short of someone saying “oh, yeah, we dropped a bottle of bat coronavirus back in October 2019″…)

          My understanding is that people have been expecting something like this to happen someday since SARS made us aware that these kinds of coronaviruses existed, so why is it surprising that it did happen?

        • Zoonoses happen *all the time*, why even look for another explanation?

          So does escape of viruses from BSL-4 labs. In fact it happens more often.

        • >>So does escape of viruses from BSL-4 labs.

          Not of new, previously unknown ones…

          >>In fact it happens more often.

          Not even close – if you’re comparing apples to apples. There is *no* historical example of a lab escape causing a pandemic … so if you are going to use historical examples of lab escapes you have to compare to *all* zoonotic infections, not just those that are of totally new viruses or started pandemics. In which case there are literally millions (basically every separate case of ornithosis/psittacosis, Newcastle disease, brucellosis, etc.)

        • Not of new, previously unknown ones…

          I think you just have a fundamental misunderstanding about what these labs are doing. You think the virus is just staying the same there or something.

          As the name indicates SARS2 is basically SARS.

        • >>As the name indicates SARS2 is basically SARS.

          except the mortality rate is an order of magnitude different, or more…

          I am not arguing lab escape is *impossible*. But possible pandemic coronaviruses has been something scientists have been watching out for – so when something entirely expected happens, why should we look for unusual explanations rather than the expected one?

        • except the mortality rate is an order of magnitude different, or more…

          If they had stopped putting SARS patients on ventilators right away and did extensive testing to dilute the CFR the mortality rate would likely be similar.

        • I don’t think the testing can be relevant here, if there were a ton of missed SARS cases to dilute the CFR it wouldn’t have died out.

          I think the difference is exactly that SARS-COV-2 is usually mild, often so mild it’s missed, whereas most SARS cases were pretty serious.

          Arguably there is somewhat of a transmissibility / deadliness tradeoff here (for the same reason Ebola etc. have not spread effectively outside their region of origin — people aren’t walking around looking fine spreading them).

        • I don’t think the testing can be relevant here, if there were a ton of missed SARS cases to dilute the CFR it wouldn’t have died out.

          Here is the case definition for SARS:

          https://www.cdc.gov/mmwr/preview/mmwrhtml/mm5249a2.htm#box

          If you were not symptomatic you did not have SARS. If the symptoms could be explained some other way you did not have SARS. If you did not have detectable antibodies a month later you did not have SARS.

        • Three of 400 healthy blood donors who donated during the SARS outbreak and one of 131 non-pneumonic paediatric inpatients were positive for IgG antibodies, confirmed by the two western-blot assays (total, 0·48% of our study population).

          […]

          However, extensive seroprevalence studies and mass screening for detection of subclinical and non-pneumonic infections are still lacking.

          […]

          The difference between the rate of non-pneumonic SARS-CoV infections in our study population and the rate of SARS in Hong Kong suggests that non-pneumonic infections are more common than SARS-CoV pneumonia and may explain cases of SARS-CoV pneumonia in patients who had no obvious contact with other patients with SARS.

          https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7112439/

          So there could have been up to 30k cases of SARS in Hong Kong alone based on this. That is vs 1755 reported cases and 298 deaths:
          https://www.who.int/csr/sars/country/2003_07_11/en/

          Obviously it suffers from all the usual selection bias, waning, and assay error problems. But the fact is you cannot compare CFRs to covid since the numbers were arrived at in completely different ways.

          Early on the US the CFR was closer to 10%, just like SARS. The testing strategy was more similar to SARS during that time as well.

        • Obvious to you; who are privy to *all* the evidence, Mr. or Miss Oz, all the evidence in-hand behind the curtain?

          While you’re on a roll why don’t you say that, because there’s “no legitimate evidence” it wasn’t produced by Martians beneath the surface of the Martian Moon Phobos, then that cannot be ruled-out; and the fact that we’re not talking about those martians is because the matter has been “suppressed” ?

          Closer to home, is the fact that no-one’s mentioned Fort Dietrick evidence that the possibility it came from *there* is also being suppressed? Shall we suppose it did *not* come from Vozrozhdeniya Island? Or is the fact that no one’s brought that up in this screed yet also evidence that another unprovable negative needs to be considered?

          And the fact that no one’s said a word about Vitamin C — well that’s evidence too, that they’re hiding something; that they’re too stupid for words; but smart enough, evidently, to participate in this vast conspiracy of suppression. Suppression of this fact; of that fact; oh Lord! See how all the pieces have gone missing; who has stolen the pieces of that jigsaw puzzle? What? You say, the pieces are right under our very noses? And what? We’re too much in the thrall of the unspeakably canny orthodoxy which causes us to ignore truth and to subscribe to patent absurdity; such authority so canny and malign it can only by an auxiliary or fifth-column of … of whom? Not Lucifer himself? No. Much, much darker than even him: The high-church of the academy.

        • I think the Fort Dietrik angle should be investigated too. And the possible link to the 2019 vaping illness.

          Seems the standard response is to just assume whatever is most politically convenient rather than use science though.

          Assumptions and hope are not science, it is faith. And that is what people are using for many different aspects of this entire fiasco.

        • >>Why did swine flu stop spreading?

          It didn’t; the H1N1-pdm09 virus is still circulating – it’s just no longer *pandemic* – it’s an ordinary seasonal flu virus. This is what happened with past flu pandemics as well.

          Pandemics end, as do sub-pandemic outbreaks (e.g. Ebola) but the pathogen *completely disappearing/going extinct* is much more unusual. Even bubonic plague is still out there, just no longer pandemic.

          There are a few other cases of diseases that seem to have outright disappeared (eg the 17th century “English sweat”) but it is not usual.

      • This has been done in China (some version of the CCR5 HIV-resistance mutation, I think, maybe not identical to the natural one) and I believe the biologist in charge was convicted of something.

        • I think so too. I wonder if a sample of the genome of those genetically modified babies would look any different than any non-modified sample… one wonders…

        • I have seen only general/non-scientific media articles on this, which I don’t trust for scientific details, but *assuming* they were correct that all that was changed was that one gene, I’m not sure how one could tell – even if it’s genuinely different than the known natural mutation (CCR5-delta32 I believe) how could you know that it wasn’t a different natural mutation?

          Something that was changed more extensively, OTOH, that could be quite different.

          But I think there’s a critical difference here. In the absence of actual evidence of genetic engineering being done (which we do have in the human case, but not the SARS-COV-2 case) there is no reason to postulate it since we know natural mutation is incredibly common.

          It’s not that engineering is *impossible* or *disproven* just counter-Occamian.

        • https://towardsdatascience.com/stop-using-the-occams-razor-principle-7281d143f9e6

          let’s try to argue without Occam’s razor in mind. it makes the world a better place. but i agree that is a difficult question.

          maybe we have to think in terms of statistics. how does this even look like? could we even rule out the option? is there a test for genetically modified? (modified in some direction perhaps, as in looking-like a natural mutation instead of just a crazy never seen before natural mutation, and perhaps here words such as natural start to lose meaning because maybe natural is what we have seen before, but not what we havent? so who knows?)

          The biggest problem is that i have the impression that we just cannot rule out this option given the current stage of genetic advances in technology (but im no expert on this, and have no idea how such tests actually look like, but is just my impression from seeing supermarket veggies and fruits, or even animals, that are genetically modified and yet look exactly like what we would see under “natural” conditions). how could we spot the difference?

        • I am not sure you can rule it out really… but I don’t think there is a reason to bring up the possibility unless there is something leaning against the normal, natural process of zoonosis.

          There are things that would show that something *is* genetically engineered (green fluorescent protein gene in a mammal, for example, or other commonly-used things like that) but I am not sure it can ever be *ruled out* because you could duplicate a natural mutation.

  3. Let’s see, a 193 page non-peer reviewed document from a biochemist previously specializing in breast cancer who is now selling a nasal spray that supposedly cures Covid19…

    Well, my Bayesian analysis puts quite a lot of prior weight on “this is bullshit”.

    • Bayesian analysis…you can’t be serious. This isn’t science, it’s business. Since 2006, this guy’s been trying, unsuccessfully, to market nasal sprays to treat everything under the sun – osteoporosis, diabetes, B12 deficiency, obesity, erectile dysfunction (see Nastech Pharmaceuticals). This time it’s a nasal spray to treat a global pandemic.

      I do have to hand it to him, though. This ridiculous “paper” will endear him to millions of conspiracy theorists who will gratefully open their wallets to such a wonderful provider of affirmation. As far as the true origin of SARS-CoV2 goes, it’s unlikely any of us will ever know for certain unless a zoonotic pathway is identified. And right now, who really cares? We’ve got a pandemic to deal with.

  4. Andrew –

    > A Bayesian analysis concludes beyond a reasonable doubt…

    I was under the impression that a Bayesian analysis wouldn’t conclude ANYTHING “beyond a reasonable doubt?” Am I wrong about that?

    • I should say, my understanding is that assuming a particular set of priors a Bayesian analysis might suggest a conclusion beyond any reasonable doubt – but that then the question is left unresolved as to whether that choice if priors is valid beyond any reasonable doubt.

      Am I wrong?

      • Suppose you are trying to determine if a coin is double headed or not… You flip the coin and get heads… Inconclusive… You flip again, heads.. inconclusive… Do this 3 more times… You’ve got 5 heads in a row… Evidence is moderate that you have a double headed coin. Then you flip and get tails… Evidence is conclusive that you do not have a double headed coin.

  5. If an unreviewed claim by some rando falls right into a common misinformation narrative and you are unable to substantively evaluate it, it might be better to just not post about it. These things feed on the oxygen of attention and is perhaps even more effective if the “scientific establishment” hates on it.

    • Ian:

      I was thinking about that, but one of the people who contacted me asking about this study seemed to be asking in an official capacity, so that I figured that this particular cat was already out of the bag.

  6. As loath as I am to link a tweet as a citation, chatter among epidemiologists that I’ve seen has basically landed on: it’s not crazy to think this was a lab release. It’s not crazy to think it’s zoonotic. Unless China suddenly discovers a strong passion for transparency, it’s not super likely we’re ever going to find out for sure. This particular person still sounds like a loon.
    https://twitter.com/CT_Bergstrom/status/1350772867613425664

  7. I don’t see why it couldn’t be both, zoonotic and accidentally escaped from the lab.

    Bio-safety BSL-4 level facilities worldwide (the first Chinese one is in Wuhan):

    https://en.wikipedia.org/wiki/Biosafety_level#Biosafety_level_4

    Interview (Indian news outlet) with F. Boyle, the gentleman who helped draft the U.S. domestic implementing legislation for the Biological Weapons Convention, known as the Biological Weapons Anti-Terrorism Act of 1989. It was approved unanimously by both Houses of the U.S. Congress and signed into law by President George H.W. Bush:

    https://www.youtube.com/watch?v=s56zAchvk3w

    The legislation:
    https://www.congress.gov/bill/101st-congress/senate-bill/993

    The paper about Boyle’s claim:
    https://www.heraldopenaccess.us/openaccess/coronavirus-is-a-biological-warfare-weapon

    I don’t think the origin matters. It is out there and the only relevant thing now is how to deal with it.

    • It’s *certainly* not an intentional bioweapon; a weapon that overwhelmingly affects the elderly while mostly sparing healthy young adults (e.g. soldiers) just doesn’t make any sense.

      Nor does something this uncontrollable make any sense as a weapon.

      • Right.
        His point is that it is certainly not intentional. However, level four labs do exist, as does the research on bio-weapons. Cover-ups have been around as long as all of humanity. It’s not sci-fi. All it takes is one accident. I believe it was what he meant.
        After all, he’s the one who drafted that legislation for Busch.

        • From the Skopec article: “He believes the virus is potentially lethal and an offensive biological warfare weapon or dual-use biowarfare weapons agent…”
          Where do you get ‘certainly not intentional’?

        • That is crazier IMO.

          Lab *escape* (of an existing virus) is IMO not an inherently ridiculous concept – but it is very counter-Occamian since natural zoonoses are common, so there’s no reason to postulate it.

          SARS-COV-2 as an intentionally engineered biological weapon *is* ridiculous; it’s not very dangerous to military populations (young healthy adults).

    • > I don’t think the origin matters.

      The origin matters in the sense of ‘what do we do to prevent outbreaks like this?’

      If it was a natural zoonosis, we need a tighter regulation of the wildlife market.

      If it was an accidental leak from a lab, we need better lab security.

      Of course, you could say ‘why not both?’ but our resources are always limited and assigning them efficiently is important.

      Finally, if it was an intentional leak, well, ‘confused’ lays down the argument for why it does not make sense as a weapon and the initially bungled Chinese response also suggests this is extremely unlikely.

      • Fred,
        I meant that the virus itself is of natural origin (not synthesized in the lab from scratch, if that’s even possible). We are no competition to mother nature, by a long shot.

        However,it is not a stretch to go with Boyle’s theory. Simply, someone was working on it, studying it and it got out.

        I know that it’s convenient to blame the wet markets, but let’s be honest and recognize that those appear everywhere in developing world and have been around for a while. Every second of every minute, there are vast numbers of mutations and interactions b/w humans and animals happening.

        Somehow, it just happened at that one place and then nothing for an entire year. I mean, nowhere else. Possible, true, but unlikely. Many parts of Africa have a lot more unregulated wet markets than China.

        In any case, all this doesn’t mean one cannot get vaccinated and protected from the infection. The fact the virus is not as bad as MERS with 30% plus mortality, doesn’t mean it didn’t originate in lab, and we can dismiss such theories. It just means it may have happened accidentally, but a lot was at stake so it had to be covered up.

        • I mean, it could theoretically be a lab escape of a natural-origin virus.

          But why is a lab more plausible/likely than any other source of human/animal interaction? It doesn’t *have* to be wet markets, there are all sorts of other human-wildlife interactions happening. Most infectious diseases probably jumped from animals at one point.

          If it’s still thought to have gone through pangolins (was there a “final” answer on bat -> human vs. bat -> pangolin -> human?) I think that makes the market explanation a lot more likely, though, since pangolins are *heavily* exploited for the wildlife trade.

  8. Apart from the actual claim it was making, I am wondering how can any statistical method tell whether or not a virus is lab-made. In essence, statistics is useful for explaining effect-of-cause, but rarely helpful for justifying cause-of-effect.

  9. A Bayesian analysis with numerical precision is silly, but the core argument behind lab-escape is a Bayesian one and seems solid.

    The prior probability of lab-escape could be on par with zoonosis. History says zoonosis for a virus this bad should be p~0.01 (based on a big bad flu 100 years ago). Lab-escape is pretty uncertain (presumably zero previous major outbreaks) but on the order of p~0.01 doesn’t seem crazy.

    But the likelihoods look radically different. p(outbreak in Wuhan | a virus escaped some bats with coronavirus hundreds to thousands of miles away) << p(outbreak in Wuhan | a virus escaped the Wuhan virology institute that was making coronaviruses super bad in order to learn about super bad coronaviruses).

    I totally dismissed lab-outbreak hypothesis early on because it just seemed like a dumb Trumpy thing, and deliberate release is still a stupid conspiracy. But having now read more credible sources on the issue, I think "some lab techs F'd up" needs to be on the table. If nothing else we need to change the way we are doing things safety-wise as lab leak is what happened, because based on known research and safety practices it is definitely something that could have happened and will happen again, just as Able Archer and Stanislav Petrov and Vladimir Arkhipov are good evidence that we needed to (and still need to) be more careful with nuclear weapons.

    • >>The prior probability of lab-escape could be on par with zoonosis.

      I would strongly disagree here, because…

      >> History says zoonosis for a virus this bad should be p~0.01 (based on a big bad flu 100 years ago).

      I don’t think so. We don’t know final severity of this pandemic since it’s not over yet, but it is almost certain to be far closer to 1957 and 1968 pandemics than 1918 (quite possibly slightly worse than the former two, but nowhere near 1918 territory*).

      And I see no reason to include the 1918 pandemic as a historical example but not 1890s one (which may actually have been a coronavirus).

      And – while it’s a different transmission mode of course – HIV is also a recent pandemic zoonosis.

      And then there’s all the ones that are serious but didn’t go pandemic, like Ebola, Marburg, etc.

      *1918 was 18 to 100 million (CDC estimates 50 million) out of a world population of about 1.9 billion, or 1% to 5%+ of the total human population — given current world population, that would be ~75 to ~400 million deaths.

      Currently reported deaths, according to the JHU tracker, are about 2.2 million.

      1957 was 1-4 million per CDC estimate, out of a world population a bit under 3 billion (0.033% to 0.133% of world population) – equivalent to something like 2.6 to 10 million from current world population. Given that vaccination programs are rapidly expanding, that range doesn’t seem unreasonable.

    • Also, how sure are we that Wuhan was actually the location of human-animal transmission, vs. merely the first significant or first *detected* outbreak? A small number of rural farmers being infected, for example, might not necessarily have been noticed…

  10. While I would not completely dismiss the lab-escape hypothesis, I understand that a major reason for the lab being in Wuhan is that Wuhan is located relatively near a lot of coronavirus-carrying bats.

    @ confused puts it very well

    Yeah, but I wouldn’t really expect a lab to be a particularly likely route for animal-to-human transmission, in a part of the world where the wildlife trade flourishes.

    If we were talking about a region with a much higher researcher-to-total-population ratio and less wildlife trade (like the Arctic or something), maybe…

    @ Navigator
    Somehow, it just happened at that one place and then nothing for an entire year. I mean, nowhere else. Possible, true, but unlikely. Many parts of Africa have a lot more unregulated wet markets than China.

    And parts of Africa are still fighting an Ebola outbreaks that are believed to have originated from “bush meat”.

    BTW, there is no really good reason to assume the original cases came from the wet market. That seems to be where the first outbreak cluster was identified but I think the Wuhan CDC believes they may have some earlier scattered cases.

  11. > all those 1.2% and 98.8% and 0.2% and 99.8% numbers on page 6 seem completely arbitrary

    Aren’t Bayesian priors always somewhat arbitrary? Is your particular claim that these priors are excessively arbitrary? If so, how do we evaluate relative arbitrariness of priors? I guess that, itself, would be some kind of prior.

    I thought the whole point of Bayesianism was to make more clear the often arbitrary and/or questionable priors that we all start research with.

    • I just looked at the table on page 6 and that table appears to be a running Bayesian probability analysis. In other words, he starts with initial probabilities in row 1 (98.8% zoonotic), and then applies Bayes formula with each piece of evidence. For example, row 2 is evaluated on page 23 and ends with, “Since the purpose of this manuscript is to evaluate the scientific evidence concerning the origin of SARS-CoV-2 no further effort will be put into these matters. […] Likelihood from initial state is unchanged following this evidence analysis.”

      It seems given how quickly you’ve dismissed this paper that this is not a standard approach in the field; however, for some someone with an undergraduate Mathematics degree, I find the approach very straightforward.

      • > It seems given how quickly you’ve dismissed this paper that this is not a standard approach in the field; however, for some someone with an undergraduate Mathematics degree, I find the approach very straightforward.

        What? Yes, priors are always made up, but it’s not just the prior, but EVERYTHING. This is an utterly non-empirical document, there is no data in these likelihoods, just paragraphs of qualitative arguments and then an arbitrary mapping to a quantitative likelihood factor. The value add by this “bayesian analysis” is hence 0. It’s just numbers and keywords thrown in for a sense of scientific authority for what is an entirely qualitative argument. Qualitative arguments are not worthless, but dressing them up as something they’re not is just a scummy way of trying to generate undue confidence.

        Bayesian arguments entirely without data are indeed not a standard approach in the field.

    • Country:

      1. All models are based on assumptions, but we want to know where the assumptions come from and how they make sense. I don’t see any good justifications for the numbers in that paper. In that paper, it’s not just priors that are made up; the likelihood numbers are all made up also. GIGO. It would be like, ummm, if someone were trying to compute the area of a rectangular field by multiplying the length times the width, but with no measurement of either of these quantities. The math is straightforward but the result won’t make sense if the inputs don’t make sense.

      2. I’ve not “dismissed this paper”; I said that their Bayesian analysis does not make sense, but I did not comment on the biology and epidemiology claims, and I stated that those were the most important parts of the paper. See the last sentence of the above post.

      • > I don’t see any good justifications for the numbers in that paper.

        What’s one example? The only specific thing I see in your post is, “all those 1.2% and 98.8% and 0.2% and 99.8% numbers” which I mentioned are just pointers to other parts of the document. I understand you’re not interested in evaluating the biological claims, so is there another claim that you were evaluating? I gave an example of row 2 in the table which points to page 23 which concludes, with some reasoning, that that particular claim doesn’t change the prior. Is that an example you think is invalid?

        • Country:

          On page 6 of that paper they have a table with 4 different instances of 1.2%/98.8% and 19 instances of 0.2%/99.8%. These numbers seem completely arbitrary, weirdly precise given that they’re all made up. It’s the job of the authors of the paper to justify their numbers; it’s not my job as a reader to do that work for them. It’s fine for you to believe the substantive conclusions of that paper—I have no informed view on this, one way or another—but if you do, I wouldn’t recommend doing so based on those 0.2% numbers etc.

        • > they have a table with 4 different instances of 1.2%/98.8% and 19 instances of 0.2%/99.8%

          As I mentioned, he’s running through the Bayesian formula for each claim and updating his priors. Where you see the numbers unchanged, that means he concluded that his priors are not changed by that particular row. He started off with a 99.8% prior that the virus is zoonotic, which seems reasonable given what we’re told by experts.

          I agree all of those rows should have been linked to the relevant parts of the report. It’s clearly an amateur paper, but I think it’s important for the academic cathedral to not stray too far away from independent thinkers in the population. For one, that’s what leads to conspiracy theories and a distrust of academics.

          I have no opinion on the paper and haven’t read it in depth. However, I did find it to be a refreshingly simple and straightforward approach to applying the Bayesian formula, although clearly it’s unintuitive since you didn’t get it.

        • I’ve emailed the man to tell him that the first 4 rows in table 6 don’t add up to 100% and that it would help if the table had a caption that it’s an iterative, cumulative application of Bayes formula, and to add a column to reference where each claim is evaluated in the report.

          Let’s see how he responds to the criticisms.

        • Oh wait, now I feel dumb. They do all add up to 100%. Anyway, the criticisms about a table caption and references still stand.

      • You should dismiss this paper. In fact, it isn’t even a paper.

        I spent more time than I should have yesterday trying to make sense of what it was and the best I can say is maybe it’s a Screed. In fact I wouldn’t even comment about it except that I spent time trying to figure out what was going on (trying to find a through-line that would allow a person to remotely judge the claims) and I may have gotten dizzy and fallen into a well.

        A person who knows a *tiny bit* about DNA and a little bit about evolution and phylogeny, would be able to get the gist of that part of his “paper.” (but I beg of you not to try)

        Some observations:

        He references some people who he discussed it with and one person whom he sites as having “done the bayesian analysis” (and I use the word analysis extremely loosely). The guy is a biostats person with some papers which is fine but I see no evidence than he knows anything about bayesian analysis. I haven’t published a paper using bayesian analysis yet, but I’ve spent three years working on learning enough so that I can. I would have done a better job than this. Or rather, I wouldn’t have entertained thinking about this, as presented.

        He pulls in lots of random bits of “evidence” to “update his prior belief” (those aren’t sarcastic quotes those are ‘I disagree that he is actually doing these things’ quotes). There is no information or methodology or supported reasoning behind what got included, different types of evidence were used interchangeably and evidence was given equal weight (even I think multiple pieces from the same paper, used sequentially)

        He includes screen captures of figures from the main paper he is refuting, but cropping them leaving out meaningful pieces (which is easy to tell because the caption/legend is underneath referring to things that you can’t see. I thought I didn’t understand the argument or the figure so I went and looked at the actual figures from the article).

        This seems like a guy who knows about virus genome sequencing was sitting down with 4 or 5 papers and some of his own ideas and decided he thought based on that info it definitely was a lab created bioweapon (at BEST that is what happened, more likely his conclusion was first).

        Any reasonably qualified person (including me) who he asked to look at this with him would be like… “Whoa whoa whoa, keep your pants on… We have to get a bunch more information. We should see if they have their data and code available, if not maybe we could find their sources and do a similar analysis. And we need to know more about varying rates of mutation when hosts switch from one Order or Family to another (versus within Genera). And find all the competent hosts … blah blah blah…”

        And he figured that was crazy he was right he would just do it himself.

        And maybe he is right. But there is no way at all to tell. I’m terrible at writing discussions, but I could be considerably more coherent than this, if only because I always have a coherent thought process and a strategy to answer the question I have or describe the process I’m curious about.

        If someone else read this in a more top down way and knows that I’m missing something I’d love to have a map to this maze. It seemed mostly narrative with various figures from other papers (cropped to show just the parts he was discussing), criticism of the conclusions or methods of those papers and a table that gets iterated upon. I read chunks here and there trying to find a section that explained his methodology but seems like I was just left with that section describing Bayes Theorem (which was quoted from a Medium post about Bayes) in the very beginning.

        Maybe there was a section that lays out his criteria for choosing papers and how they are incorporated into his “updating”. If he’s right (based on what I read there is no way to know whether there is a 90%, a 50% chance, or a 0.1% chance he’s right.) then someone should do this in a way that is even a little bit coherent or consistent.

  12. He makes testable claims. The chief one is this:

    “The most significant evidence provided herein is the finding from RNA-Seq performed by the Wuhan Institute of Virology (WIV) of lavage patient samples collected on December 30, 2019. These ICU patients were the subject of the seminal paper, entitled, “A pneumonia outbreak associated with a new coronavirus of probable bat origin,” from Dr. Zhengli Shi and colleagues that first characterized SARS-CoV-2.

    This author has confirmed that the RNA-Seq of all five patients contained SARS-CoV-2 sequences. Surprisingly the specimens also contained the adenovirus “pShuttle” vector, developed by Chinese scientists in 2005 for SARS-CoV-1. Two immunogens were identified, the Spike Protein gene of SARS-CoV-2 and the synthetic construct H7N9 HA gene. Hundreds of perfectly homologous (150/150) raw reads suggest this is not an artefact. Reads that cross the vector-immunogen junction are identified. An example of the read contigs for CoV-2 is shown in this figure”

    He claims that he found an adenovirus based vaccine containing the SARS-CoV-2 spike protein (similar to the Oxford/Astrazeneca vaccine) in the data published with the pneumonia outbreak paper.

    It is a testable claim.

    If it is true then we don’t need to worry about Bayesian this or Bayesian that.

    If false then that does undermine his credibility. He’d have to be a liar, not just incompetent.

  13. Isn’t “Bayesian” reasoning merely a synonym for “Probabilistic” reasoning ?

    Attachment of the proper name makes it sound like a sectarian cult of some sort.

    If one accepts a certain *definition* for conditional probabilities (as the probability of a conjunction divided by that of one of the conjuncts) then all the rigamarole that follows are merely consequences of a purely conventional definition.

    Is there some variety of probabilistic inference that does *not* subscribe to the use of this aforesaid definition?

    • 1. Why don’t they publish the relevant documents themselves?
      2. The request was not granted.

      What this does show (which is already well known, e.g. via the authorship of the Chapel Hill paper) is that the WIV is internationally well connected; and that mutual visits take place.
      Generally, academic researchers like to talk about their projects with their colleagues when they visit.

      Therefore, it appears likely that if there was research at WIV that would lead to SARS-CoV-2, someone else (e.g. Peter Daszak) would also know about it.

      “British zoologist and the president of EcoHealth Alliance Peter Daszak is the only individual to be part of both the WHO and The Lancet teams investigating the origins of the Wuhan coronavirus pandemic.”

      Daszak had ample opportunity to officially find something that he knows inofficially, and thus to answer the important scientific question as to the origin of the virus. But that has not happened.

      The WIV keeps a genetic index of its viruses (part of the sequence), and the WIV checked the candidates for SARS-CoV-2 when the news of the outbreak broke, and the closest virus they had in the lab was RaTG13, which isn’t close enough to support the lab escape theory.
      (Shi Zhengli could presumably have circumvented a coverup order via her international contacts.)

      At this point, pursuing the lab escape theory is equivalent to questioning the scientific and moral integrity of virus researchers like Peter Daszak and Shi Zhengli. These people (and many more) would have had to conspire to cover up a man-made lab-escape origin of Covid-19.

      • And why do you think Peter Daszak is being honest? Have you even looked at the emails and how he was orchestrating a letter in Lancet to call any lab related hypothesis as a conspiracy theory?
        That doesn’t sound scientific at all to me.

    • “Would the existence of a verified document containing a request by a well known alliance for funding to insert a furin cleavage site between S1 and S2 alter the Baysian prior?”

      Given that the proposal (which apparently was never funded) was for this part of the research to be done at a research facility here in the United States (a Duke lab, in North Carolina), no, it would not change any prior on the probability of SARS-CoV-2 being engineered in a lab in Wuhan. Not at all.

Leave a Reply to Navigator Cancel reply

Your email address will not be published. Required fields are marked *