While I reminisce my swindles of yesteryears, a new one has been unfolding here before me. I say with pride that my student has become the master. @PeterMillerc030 😉

@KristianAndersen Ballsy move shifting the axis like that.

@KristianAndersen Massey’s on thin ice here. How does a back of the envelope adjustment in a blog post compare to cropping sequences out of a figure in a scientific manuscript that would have showed that, no, you did not identify a true intermediate genome? Because that’s what Massey et al did. Further he wrongly claimed that this single sample, were it described accurately, falsified Pekar et al 2022.

Obvious screw up not keeping back of the envelope estimates separate from other things, but what’s the significance here? The equivalent calculation in Pekar 2022 accounting for the revision to that case and ruling out artifactual

intermediate sequences shifts the edge of 95% CI from 30/Nov to 8/Dec for date of infection of the primary case, I think. Is someone somewhere in the world leaning on this figure for some reason?

A rebuttal of some mathematical claims about early Covid-19 made by Scott Alexander.

https://arguablywrong.home.blog/2024/04/09/how-likely-is-it-for-covid-to-establish-itself/

As evidence against an earlier introduction, Scott also presents a doctored version of Fig. 3E from Pekar et al.'s "Timing the SARS-CoV-2 index case in Hubei province", 2021. It seems this misleading image was used for the Rootclaim debate.

https://twitter.com/nizzaneela/status/1777989261817508165?t=iKdlIodfWEfzFVWxGlVtrA&s=19

It's only misleading if you present it without context.

Here are Peter's slides, in the document he says "Using December 10th as the first ascertained case makes it most likely that the first case is late November or early December."

@Odoacre He presented as if it were unaltered from Pekar et al (2021), no? I'm not sure how that isn't misleading even if inadvertent. It's also wrong as set out in this thread.

https://twitter.com/nizzaneela/status/1777996139385491577?t=7YlIZWk6f1y8YnCGTm86rQ&s=19

Further on how you can't simply adjust it 10 days. Since this of course another paper has come out suggesting spillover was Aug-Oct 2019 which is closer to the estimates of Caraballo-Ortiz, Kumar and Bloom.

https://twitter.com/Dissenting2020/status/1778352829356188094?t=Ex9dei9g5ZnYCJ_3DlStxQ&s=19

@MikePa67d And don't get me started on not mentioning the difference between the 91Q and ZY6 mutations, how can we ever trust anything Peter and Scott said?

@DavidKochanov Personally I only trust obscure Nepalese journals that nobody except me has ever heard of.

Indeed. Zoonotic spillover advocate Bob Garry just had a paper retracted for image manipulation. No one thinks that undermines his other work. But it's worth correcting.

@Odoacre I looked up the blog post where I first used this. It literally says that it's an adapted figure!

Though I will agree it's not as clear in the debate slides where I pared down all the text.

I also looked up how Rootclaim responded to my slide (their text in yellow):

He responded the same way to every slide where I put some date or some model that estimated dates. And he would have replied the same way to literally any model I wrote down.

That's right, Saar doesn't care when the pandemic started, he's just got a gut feeling that it was probably a lab leak. I would have loved spending more time discussing epidemic models, that would have been an easy conversation to win.

@PeterMillerc030 thanks for the clarification Peter.

@PeterMillerc030 "We don't make any claims based on dates" isn't particularly accurate with Rootclaim appealing to dates associated with all sorts of things. Is selectively throwing out useful data a precondition of the terms of a new rootclaim debate?

@zcoli Dates are like probabilities. Saar thinks they're only worth paying attention to if they support his theory.

@PeterMillerc030 Dates in tabloids are data. Dates attached to patient samples that are corroborated in many ways can't be trusted.

@zcoli The real problem is that we can't trust complicated models. That's why it's key to get your arguments from the simplest sources possible, and the Daily Mail is an excellent source, much easier to read than any scientific paper.

Scott has a new post with responses to his responses, and concludes with:

If it helps, I’m currently working out terms for a 6-digit lab leak bet of my own (no guarantee this will come to fruition, most of these fall apart in the resolution criteria stage). I feel bad for not being willing to answer every possible lab leak argument going forward, but hopefully offering lab leakers a few hundred thousand dollars if I’m wrong will be a suitable consolation prize.

For now, I’m still at 90-10 zoonosis.

Let's bet on this.

@Joshua I read that to be exclusively a bet, like "I bet 100k that there won't be a consensus for zoonosis by 2030" or something, rather than a bet about winning a certain debate. does he mention a debate related to the bet?

@Bayesian I think he means a bet in the same way that Miller and Wilf bet $100k, and the bet was then settled by judges that heard their arguments.

You can bet on no debate happening in the MC market I made though!

@Joshua The post doesn't even touch on many strong arguments against Rootclaim's case as it stands today. Here is a 12-nucleotide "clean insert" in SARS2 with CGG-CGG-CGG encoding R-R-R. If my math is right this wipes out Rootclaim arguments that are 20:1 ("12 nucleotides clean insertion") and 5.3:1 ("Insertion is from a foreign source") in favor of lab leak. This is hardly the only example with these sorts of features and for every one that's observed in sequencing data there are countless ones that aren't observed.

https://cov-spectrum.org/explore/World/AllSamples/Past6M/variants?variantQuery=ins_22204%3ATGGCGGCGGCGG

Sticking at 90:10 is a good way to entice someone to take the other side of the bet, though.

In the Rootclaim debate it was argued SARS-CoV-2 spilled over in November 2019. New paper suggests spillover between August and early October 2019. This presumably makes the December 2019 cases Worobey et al focused on less informative.

The authors found "that the presence of a statistically robust clade in the phylogenies of gene S and RBD, including two BANAL and two Guangdong pangolin coronaviruses and closely located to SARS-CoV-2, is most probably due to the horizontal gene transfer of gene S from BANALs to Guangdong pangolin coronaviruses that occurred in the middle of 2018". Unmentioned is that "horizontal gene transfer" - can occur in the wild (rarely) or can be induced in he lab artificially "transgenic".

Whatever the case it looks like the answer is outside the December cases in the WHO report.

https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0301195

Previous studies using phylogenetic inference and epidemiologic simulations, focusing on the SARS-CoV-2 lineage A and B, have identified the probable spillover time being between late October 2019 and mid-December 2019 [37]. Interestingly, the officially accepted SARS-CoV-2 emergence data early of December 2019 SARS-CoV-2 [38] falls just outside of our 95% highest posterior density (HPD) confidence intervals (see Fig 6).

This is a quotation from your link. What is this nonsense?

Ignoring other issues with this manuscript, Figure 6 shows their estimated time for the most recent common ancestor, reference 37 is an estimate of the range of plausible dates of first human infections, and reference 38 is the first reported date of symptom onset. These are three different things.

Promoting the latest shiny thing based on whether or not it supports your favored hypothesis regardless of being unsound only shows how little evidence there actually is to support the hypothesis.

@zcoli I like your approach much better than Peter's. Pick up on a minor inconsequential point and make a mountain out of a mole hill.

While avoiding any attention to the actual scandal of Pekar 2022 in which multiple significant errors were discovered. A wise man once said, "Always accuse your enemies of your own sins"

@KristianAndersen the multiple spillover theory of Pekar et al seemed far fetched from the outset with lineage A and B only 2 mutations apart and incomplete case ascertainment. The errors identified since on pubpeer, evidence of sampling bias and intermediate genomes have just added to that.

https://twitter.com/Biol4Ever/status/1634981519252127746?t=CFh34FYouRXqbUnvega-Lw&s=19

@MikePa67d Are you changing the subject after embracing the latest shiny thing of a paper because it had a sentence in it that you like even if it wasn’t supported well at all?

@zcoli it's their main conclusion which appears consistent with Caraballo-Ortiz (2022), Kumar 2021 and Bloom 2021. It's the latest of several papers and analyses that dispute Worobey et al and Pekar et al. It looks like there was a single introduction in humans. Emergence in humans is concurrent with the MRCA.

@MikePa67d None of those three papers consider collection dates in constructing phylogenetic trees so you’re gonna have to be clear what “it’s their main conclusion” is referring to here because this is something totally different as are its conclusions.

Edit: and of course upending Pekar 2022 and other papers with similar estimates is not even a conclusion of this paper — it plainly concludes that its tMRCA ranges are consistent with previous estimates.

@MikePa67d > Emergence in humans is concurrent with the MRCA.

@DavidKochanov well the TMRCA may post date the spillover if early variants go undetected. Does this predate or post date emergence in humans? Depends if there was only a single spillover, and it seems there was if Lv et al are right.

@MikePa67d You just copy-pasted some tweet and the guy who wrote the original tweet( https://twitter.com/Dissenting2020/status/1777600237361639576) somehow thinks that the TMRCA from the plos paper and the TMRCA from this paper - https://academic.oup.com/ve/article/10/1/veae020/7619252?login=false are the same.

@DavidKochanov sorry I should have just linked their tweet. I took it that they were referring to Lv simply because it's the most up to date evidence concerning whether there was a single point of emergence or multiple spillovers with the new sequences.

@MikePa67d The earliest "A0" sequence in Lv is a sublineage already sequenced and reported in Wuhan and Shanghai. Doesn't push back "A0" origin date at all and is consistent with emergence in December. The "A0"-defining mutation is one of the most common across the whole SARS2 genome. Totally expected to see a couple more in Shanghai with this many new sequences.

@DavidKochanov

If it's a single spillover, the date of spillover will be BEFORE the TMRCA. This is important because even I'm not sure about the double spillover hypothesis after Lv et al.

https://bsky.app/profile/kgandersen.bsky.social/post/3kn4jizql2t2y

Not wise to play gotchas with something that calls into question the market theory.

While it's likely you're just parroting what my colleagues are saying on X, they are not the smartest bunch. We've worked really hard to pull off that illusion, let's make sure it stays that way.