[tripletao]

Their terminology would indeed seem wrong, but the WIV has apparently used a similar assembly strategy (leaving the sites in the final assembly) before:

https://twitter.com/jbkinney/status/1583267221047869441

I'd worry about their false discovery rate, for the same reason I worry about the large number of parameters in Pekar's epi model. It's still an interesting result, though.

[mikeyouse]

As far as I've been able to read, those sites aren't in the final assembly in Shi's papers - just in the primers. And even more worrying than the false discovery rate are all of the missing genomes they should be comparing. I fear they've left them off because they punch big ole holes in the theory... e.g.

https://twitter.com/zhihuachen/status/1583258714340892672

[lamontcg]

Another good twitter thread on it this morning:

https://twitter.com/acritschristoph/status/15834864034169692...

This is the real kicker to me:

> What about missing sites? The authors propose that someone made a bizarre combination of additions and deletions of cut sites. RecCA matches SARS-CoV-2 at all missing sites because other viruses do. E.g. this one: similar to RpYN06 not just at the mutation, but the entire region.

Clear evidence of recombination across the whole region and not just mutations to manipulate the cut site.

And Francois Balloux seems to have deleted his twitter account this morning.

[tripletao]

As far as I can tell, WIV (and other) researchers have left other restriction sites in final genomes, but not BsaI and BsmBI sites. That makes sense given the typical use of those two enzymes. So I'm inclined to agree there's nothing obviously special about that combination of restriction sites and spacings, which increases my concern that this is just a false discovery.

Jesse Bloom said he'd try reproducing with a wider range of natural viruses and potential synthetic assembly strategies. Unless and until that still gets an interesting p value, I'd agree this is oversold.

[tripletao]

> As far as I've been able to read, those sites aren't in the final assembly in Shi's papers - just in the primers.

Where did you read that? Kinney explicitly asserts the opposite, but I can't figure out what he's referring to. Someone else linked to Figure S9 from their 2017 PLOS Pathogens paper, which indeed seems not to contain the site in the final assembly. I've edited my other comment here to reflect that.

[mikeyouse]

Nope - the restriction sites aren't in the final assembly in that paper. These authors just don't understand the tools remotely well enough to have written their paper. It will never be published.

https://twitter.com/alchemytoday/status/1583361758903013376

[tripletao]

Yeah, I think Kinney is wrong here. Someone else found a survey paper noting the theoretical possibility of using BsaI and BsmBI in a way that leaves the sites in the final genome, but I haven't seen anyone who actually did that. So I'm leaning to a false discovery here.