How likely is it that SARS-CoV-2 originated in a laboratory?

post by CTVKenney · 2021-01-25T20:22:16.105Z · LW · GW · 27 comments

This is a question post.

I read this article in NY magazine, on the coronavirus lab-leak hypothesis. I found it fairly convincing, so I think the chance that SARS-CoV-2 passed through some human laboratory before spreading to people is better than not. But when I tried following up by reading other sources, I found myself falling into the trap of trying to just confirm my existing belief. How might we estimate a probability that SARS-CoV-2 passed through a human laboratory at some point before spreading to humans?


answer by landfish · 2021-01-25T21:00:58.794Z · LW(p) · GW(p)

I've done over 200 hours of research on this topic and have read basically all the sources the article cites. That said, I don't agree with all of the claims. I do not think the SARS-CoV-2 virus is very likely to have been created using the RATG13 virus, because of the genetic differences spread out throughout the genomes. However, there are many other paths that could have led to a lab escape, and I'm somewhat agnostic between several of them.

I don't have a lot of time to investigate this further, but if someone was going to spend serious time on it, then I'd be happy have several calls with them, discuss sources & share my notes with them.  At this point I think a lab leak is more likely than not, with the strongest piece of evidence being the confluence of the location of the first known outbreak + location of the world's top lab studying SARS-like coronaviruses + absence of related viruses detected nearby + absence of evidence of any other plausible origin.

I highly recommend following Alina Chan on Twitter, who done a lot of interesting work on this question & has appeared to me to be pretty discerning.

If I were going to spend a bunch more time on this, I'd try to conduct an estimate using a Bayesian model, probably starting here: and creating my own estimates for each claim + writing out arguments for why.

comment by landfish (jeff-ladish) · 2021-01-26T01:14:23.387Z · LW(p) · GW(p)

A random observation I want to note here is the relative lack of good disagreement I've seen around questions of SARS-CoV-2 origin. I've mostly seen people arguing past each other or trying to immediately dismiss each other. This seems true of experts in the space in addition to non-experts. I'd love to see better structured disagreement, i.e. back and forth in journals or other public forums. This might be a good topic for adversarial collaboration.

comment by landfish (jeff-ladish) · 2021-01-25T21:25:33.242Z · LW(p) · GW(p)

For context, I have a background in evolutionary theory (though nothing specific to viruses or pathogens) and have recently transitioned from part time to full time research in the longtermist biosecurity space.

When investigating this question, I found researcher's arguments pretty easy to follow, but found some of the claims about ease of engineering to be hard to follow because they often relied on tacit knowledge like "how hard / expensive is it make an infectious clone of a new coronavirus".  And some the more technical molecular phylogenetics were difficult as well (what can we infer from dN/dS of various parts of the SARS-CoV-2 vs. RATG13 genomes, and how does selection for codon preference influence this analysis). I'd love to talk with someone who feels like they have a good grasp of either of these areas.

Replies from: spacecadet, gerald-monroe
comment by spacecadet · 2021-01-26T15:55:58.204Z · LW(p) · GW(p)

There are a group of researchers concerned with CoV19 origins who frequent Twitter and use the moniker #DRASTIC. They count a number of geneticists / microbiologists in their number. See this list:

@ydeigin, @__ice9, @MonaRahalkar, @Rossana38510044, @Ayjchan and @AntGDuarte may be good candidates for your questions. 

Note that they consider RATG13 to be a chimera designed to obfuscate research.

comment by Gerald Monroe (gerald-monroe) · 2021-01-26T03:40:30.801Z · LW(p) · GW(p)

Ok, so here is what I read (my knowledge may be out of date):

 There is a way to set up a chain of animals in a "gain of function" experiment.  You start with the wild-type virus, and infect the first animal.  You put a gap to the next animal where the viral particles able to bridge that gap (out of the very large number of copies some mutated in the infected animal) are more probable to be capable of bridging gaps.

Eventually in the last animal, the gap is a large air gap, and the virus is now airborne in lab animals.  All it takes for a leak after that is a single mistake by a laboratory employee - such as a faulty seal or air filter or procedure error - and they become infected.  They then spread it to someone in Wuhan and that's your outbreak.

This same 'gap bridging' is what is creating these mutant variants of Covid that are more infectious.  

Anyways this chain of events can occur 'naturally', in the same way that enough U-235 can concentrate itself naturally to form a nuclear reactor, it just isn't very likely.

Part of this is the exactness of the setup.  In a laboratory gain of function experiment, you carefully control each gap to force the virus to evolve to bridge it. (each gap is a little bit more difficult in a smooth progression)  In some random cave or mineshaft, conditions are chaotic and the virus is not under as much pressure.  Sort of like the difference between using pure neutron reflectors and pure uranium ore to make a nuclear pile and relying on nature to make one by accident.

Replies from: Richard_Kennaway
comment by Richard_Kennaway · 2021-01-26T11:59:28.083Z · LW(p) · GW(p)

Why do people do this sort of experiment? Why do they deliberately try to create infectious plagues? Is there some real, useful knowledge that comes out of such experiments, or do they just want to see the world burn?

Replies from: gerald-monroe, spacecadet
comment by Gerald Monroe (gerald-monroe) · 2021-01-26T20:49:04.911Z · LW(p) · GW(p)

Well it's hard to study something in theory.  Making a plague - and using a procedure that could happen in nature it's just improbable - allows for these scientists to play around with it and see how it works.

I think it's research that should be teleoperated labs in the middle of a hot desert or other wasteland free of hosts to spread it...

Replies from: Richard_Kennaway
comment by Richard_Kennaway · 2021-01-26T21:36:38.749Z · LW(p) · GW(p)

Viruses are known to sometimes jump from one species to another, so what is different about how they are encouraged to do so in a lab? Are they doing anything but forcing it to happen faster?

I am not a biologist and this is wild speculation, but I wonder if, when you force pathogens to pass from one host species to another to another to another, are you selecting them not just for making each individual jump, but for being able to jump from any species to any other more easily? You would then end up with a meta-infective virus able to mutate fast enough to spread through the whole animal kingdom and outpace any attempts at vaccination.

Replies from: gerald-monroe
comment by Gerald Monroe (gerald-monroe) · 2021-01-26T21:56:30.332Z · LW(p) · GW(p)

They are forcing it to happen a lot faster, yes.

Fortunately such a meta virus that can kill everything is probably difficult to produce from the set of amino acids shared in common to life on earth. (This is essentially what a gray goo scenario is - life on earth is probably not maximally efficient and it is probably possible to build full synthetic organisms on the scale of cells that self replicate with optimized internal components)

comment by spacecadet · 2021-01-26T15:20:19.995Z · LW(p) · GW(p)

GoF research assists with vaccine development. Apparently that was the reasoning behind GoF research conducted on bat CoVs at the Wuhan Inst. of Virology.

And the results of all these experiments were lots of novel contagions and no vaccines. Complete and utter madness.

comment by ESRogs · 2021-01-25T22:09:19.474Z · LW(p) · GW(p)

This may be a bit of a pedantic comment, but I'm a bit confused by how your comment starts:

I've done over 200 hours of research on this topic and have read basically all the sources the article cites. That said, I don't agree with all of the claims.

The "That said, ..." part seems to imply that what follows is surprising. As though the reader expects you to agree with all the claims. But isn't the default presumption that, if you've done a whole bunch of research into some controversial question, that the evidence is mixed?

In other words, when I hear, "I've done over 200 hours of research ... and have read ... all the sources", I think, "Of course you don't agree with all the claims!" And it kind of throws me off that you seem to expect your readers to think that you would agree with all the claims.

Is the presumption that someone would only spend a whole bunch of hours researching these claims if they thought they were highly likely to be true? Or that only an uncritical, conspiracy theory true believer would put in so much time into looking into it?

Replies from: jeff-ladish
comment by landfish (jeff-ladish) · 2021-01-25T23:18:57.432Z · LW(p) · GW(p)

I should have worded that better. I copied that sentence from a facebook post where I had a claim above that sentence that said something like, "I think this article is basically correct in its interpretation of the literature". The disagreement is about claims the  NY mag article made that weren't backed up by sources / were the authors original speculation. I meant to convey "I think the NY article did a decent summarization of the articles he cited -- that being said, while I agree with the general thrust of the article, I think there some points the author speculated about that are likely wrong"

Replies from: ESRogs
comment by ESRogs · 2021-01-26T00:25:54.735Z · LW(p) · GW(p)

Got it, thanks for the clarification.

comment by CTVKenney · 2021-01-28T23:54:27.071Z · LW(p) · GW(p)

With regard to the rootclaim link, I agree that it would be good to try to adapt what they've done to our own beliefs. However, I want to urge some caution with regard to the actual calculation shown on that website. The event to which they give a whopping 81% probability, "the virus was developed during gain-of-function research and was released by accident," is a conjunction of two independent theses. We have to be very cautious about such statements, as pointed out in the Rationality A-Z, here

comment by River (frank-bellamy) · 2021-01-25T21:54:49.849Z · LW(p) · GW(p)

How do you reconcile the hypothesis that it escaped from a lab in China with the reports that covid-19 antibodies were found in more than a dozen blood samples taken in Italy in early October 2019, and therefor must have been circulating in Italy in September 2019?

Replies from: jeff-ladish, Daniel V, stuart-anderson
comment by landfish (jeff-ladish) · 2021-01-26T00:49:15.919Z · LW(p) · GW(p)

There have even been claims of SARS-CoV-2 in March 2019, which I think are almost certainly false positives:

comment by Daniel V · 2021-01-25T22:48:49.835Z · LW(p) · GW(p)
  1. Escaped/circulated earlier than officially reported.
  2. False positives.
Replies from: jeff-ladish
comment by landfish (jeff-ladish) · 2021-01-26T00:07:26.166Z · LW(p) · GW(p)

Yep, these are the two hypotheses. So far I think 2) is a lot more likely. Decent thread on it here: (or thread reader app link: )

comment by Stuart Anderson (stuart-anderson) · 2021-01-26T04:43:59.601Z · LW(p) · GW(p)


Replies from: Viliam
comment by Viliam · 2021-02-02T19:01:31.318Z · LW(p) · GW(p)

Don't many other countries do similar things? Even individuals often order stuff from China.

Replies from: stuart-anderson
comment by Richard Dixon (richard-dixon) · 2021-05-06T10:25:14.531Z · LW(p) · GW(p)

This is fairly convincing that it's a plausible and even likely:

answer by Stephen Reed · 2021-01-30T04:21:43.034Z · LW(p) · GW(p)

Here is the link to a recent relatively thorough article using Bayesian analysis to argue for a laboratory release of the virus.


A Bayesian analysis concludes beyond a reasonable doubt that SARS-CoV-2
is not a natural zoonosis but instead is laboratory derived


Comments sorted by top scores.

comment by kithpendragon · 2021-01-26T17:08:14.303Z · LW(p) · GW(p)

In this context, I'm an ordinary civilian with no specialized knowledge of genetics or virology. I read the linked article, and saw an interesting story built off relevant history and lots of conjecture. I've also read the comments up to this point. Here are the arguments as I see them right now:

SARS-CoV-2 (hereafter "the Virus") must have an origin. Viral origins can be natural or artificial. Natural viruses evolve from other strains, sometimes in animals. The Virus shares a lot of characteristics with a certain bat virus, which is evidence it evolved in bats and then transferred to humans. It is therefore (at least originally) natural. Nobody seems to have any issues up to this point.

The question of how the Virus got to humans has a number of possible answers that seem to boil down to:

  • A random mutation in a bat virus may have caused it to be infectious to humans.
  • The Virus may have been engineered to be infectious to humans for legitimate or nefarious reasons.
    • The engineering may have been by design
    • The engineering may have been accidental

In favor of the random mutation possibility:

  • This is a thing that happens sometimes
  • There were opportunities for humans to have been exposed to the relevant bat virus in large concentrations for extended periods
  • The Virus's genome does not show obvious signs of deliberate tampering

In favor of engineering, either accidental or deliberate:

  • The virus is suspiciously infectious to humans. There are lab conditions that would encourage this kind of mutation.
  • We know that labs at least had the opportunity to collect samples before the first noticed outbreak
  • The Wuhan Institute of Virology, which had a sample of the bat virus, is really close to the Wuhan Center for Disease Control and Prevention, which first noticed the Virus in the population.

It all seems circumstantial to me except the examination of the genome. Even the article started off admitting that there is a distinct lack of hard evidence here. If anything, the balance of evidence looks (to my eye) to be (barely) in favor of a random mutation resulting from the massive exposure of workers to the bat virus. We know that the Virus causes lots of asymptomatic infections, so I see no reason to believe that those workers didn't pass it on before they were hospitalized for their own infections. We know samples of the bat virus spent time in a lab, but the lab time doesn't seem to be necessary for the Virus to reach the general population given the presence of the workers. Given the known behavior of the Virus, I claim it could easily have moved unchecked from Mojiang to Wuhan as a barely noticed, mostly asymptomatic infection the same way we saw it spread mostly untracked through a number of other countries while we were watching for it. It then popped up in Wuhan the same way hotspots have been popping up all over the world ever since.

comment by Lantalia (GryMor) · 2021-01-26T07:46:31.795Z · LW(p) · GW(p)

To clarify, when you say 'originated in a laboratory' do you mean engineered in a laboratory; evolved in an intentionally infected lab animal; or transiently stored in a laboratory?

These are very different hypotheses, but are often conflated.

So far, there is strong evidence that it was not engineered [the spike protein is novel, and not something from the geneticists toolbox], and I haven't seen any evidence that would favor lab storage or evolution over the much larger wild populations of bats and intermediary animal hosts.

Replies from: CTVKenney
comment by CTVKenney · 2021-01-26T14:45:48.232Z · LW(p) · GW(p)

I mean to include all the alternatives that involve the virus passing through a laboratory before spreading to humans; so all the options you list are included. There's nothing wrong with asking about the probability of a composite event.