March Coronavirus Open Thread

Chinese virology researcher released something claiming it SARS-2 might even be genetically-manipulated after all? ZC45 and/or ZXC21 backbone. Claims that the RaTG13 genome was a concocted cover-up. I'm still assessing the sub-claims, but I think she makes some pretty cogent-sounding arguments for it.


(That the RaTG13 sequence could have been concocted is something I kinda wish I'd thought of earlier.)

See here for my further thoughts on this.

EDIT: After assessing, I'm not finding the GMO claims convincing. The RaTG13 story does seem to have something weird going on, and there's several people and papers that note weird inconsistencies (See the further thoughts, I don't have a simple explanation.).

Jimrandomh's Shortform

Chinese virology researcher released something claiming that SARS-2 might even be genetically-manipulated after all? After assessing, I'm not really convinced of the GMO claims, but the RaTG13 story definitely seems to have something weird going on.

Claims that the RaTG13 genome release was a cover-up (it does look like something's fishy with RaTG13, although it might be different than Yan thinks). Claims ZC45 and/or ZXC21 was the actual backbone (I'm feeling super-skeptical of this bit, but it has been hard for me to confirm either way).

https://zenodo.org/record/4028830#.X2EJo5NKj0v (aka Yan Report)

RaTG13 Looks Fishy

Looks like something fishy happened with RaTG13, although I'm not convinced that genetic modification was involved. This is an argument built on pre-prints, but they appear to offer several different lines of evidence that something weird happened here.

Simplest story (via R&B): It looks like people first sequenced this virus in 2016, under the name "BtCOV/4991", using mine samples from 2013. And for some reason, WIV re-released the sequence as "RaTG13" at a later date?

(edit: I may have just had a misunderstanding. Maybe BtCOV/4991 is the name of the virus as sequenced from miner-lungs, RaTG13 is the name of the virus as sequenced from floor droppings? But in that case, why is the "fecal" sample reading so weirdly low-bacteria? And they probably are embarrassed that it took them that long to sequence the fecal samples, and should be.)

A paper by by Indian researchers Rahalkar and Bahulikar ( https://doi.org/10.20944/preprints202005.0322.v1 ) notes that BtCoV/4991 sequenced in 2016 by the same Wuhan Virology Institute researchers (and taken from 2013 samples of a mineshaft that gave miners deadly pneumonia) was very similar, and likely the same, as RaTG13.

A preprint by Rahalkar and Bahulikar (R&B) ( doi: 10.20944/preprints202008.0205.v1 ) notes that the fraction of bacterial genomes in in the RaTG13 "fecal" sample was ABSURDLY low ("only 0.7% in contrast to 70-90% abundance in other fecal swabs from bats"). Something's weird there.

A more recent weird datapoint: A pre-print Yan referenced ( https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7337384/ ), whose finding (in graphs; it was left unclear in their wording) was indeed that a RaTG13 protein didn't competently bind their Bat ACE2 samples, but rather their Rat, Mouse, Human, and Pig ACE2. It's supposedly a horseshoe bat virus (sequenced by the Wuhan lab), so this seems hecka fishy to me.

(Sure, their bat samples weren't precisely the same species, but they tried 2 species from the same genus. SARS-2 DID bind for their R. macrotis bat sample, so it seems extra-fishy to me that RaTG13 didn't.).

((...oh. According to the R&B paper about the mineshaft, it was FILTY with rats, bats, poop, and fungus. And the CoV genome showed up in only one of ~280 samples taken. If it's like that, who the hell knew if it came from a rat or bat?))

At this point, RaTG13 is genuinely looking pretty fishy to me. It might actually take evidence of a conspiracy theory in the other direction for me to go back to neutral on that.

E-Protein Similarity? Meh.

I'm not finding the Protein-E sequence similarity super-convincing in itself, because while the logic is fine, it's very multiple-hypothesis-testing flavored.

I'm still looking into the ZC45 / ZXC21 claim, which I'm currently feeling skeptical of. Here's the paper that characterized those: doi: 10.1038/s41426-018-0155-5 . It's true that it was by people working at "Research Institute for Medicine of Nanjing Command." However, someone on twitter used BLAST on the E-protein sequence, and found a giant pile of different highly-related SARS-like coronaviruses. I'm trying to replicate that analysis using BLAST myself, and at a skim the 100% results are all more SARS-CoV-2, and the close (95%) results are damned diverse. ...I don't see ZC in them, it looks like it wasn't uploaded. Ugh. (The E-protein is only 75 amino acids long anyway. https://www.ncbi.nlm.nih.gov/protein/QIH45055.1 )

A different paper mentions extreme S2-protein similarity of early COVID-19 to ZC45 , but that protein is highly-conserved. That makes this a less surprising or meaningful result. (E was claimed to be fast-evolving, so its identicality would have been more surprising, but I couldn't confirm it.) https://doi.org/10.1080/22221751.2020.1719902


I think Yan offers a reasonable argument that a method could have been used that avoids obvious genetic-modification "stitches," instead using methods that are hard to distinguish from natural recombination events (ex: recombination in yeast). Sounds totally possible to me.

The fact that the early SARS-CoV-2 samples were already quite adapted to human ACE2 and didn't have the rapid-evolution you'd expect from a fresh zoonotic infection is something a friend of mine had previously noted, probably after reading the following paper (recommended): https://www.biorxiv.org/content/10.1101/2020.05.01.073262v1 (Zhan, Deverman, Chan). This fact does seem fishy, and had already pushed me a bit towards the "Wuhan lab adaptation & escape" theory.

Spiracular's Shortform Feed

On Community Coordinator Roles

Some things I think help are:

  • Personal fit to the particular job demands (which can include some subset of: lots of people-time, navigating conflict, frequent task-switching, personal initiative, weighing values (or performing triage), etc.)
  • Personal investment in appropriate values, and a lasting commitment to doing the role well.
  • Having it be fairly clear what they're biting off, what resources they can use, and what additional resources they can petition for. Having these things be actually fairly realistic.

(Minimum expertise to thrive is probably set by needing to find someone who will advocate for these things well enough to make them clear, or who can at least learn how to. Nebulous roles can be navigated, but it is harder, and in practice means they need to be able to survive redefining the role for themselves several times.)

  • Having some connections they (and others) respect who will help support what (and how) they're doing, lend some consistent sense of meaning to the work, and people who will offer reliable outside judgement. Quite possibly, these things come from different people.

(Public opinion is often too noisy and fickle for most human brains to learn off of it alone.)

  • For roles that require the buy-in of others, it's helpful if they have enough general respect or backing that they will mostly be treated as "legitimate" in the role. But some roles really only require a small subset of high-buy-in people.
  • Also, some sort of step-down procedure.

Extracted from a FB thread, where I was thinking about burnout around nebulously-defined community leader roles, and what preparedness and a good role would look like. Parts of this answer feel like I'm being too vague and obvious, but I thought it was worth making the list slightly more findable.

[META] Building a rationalist communication system to avoid censorship

I've seen some presentations about how to do style-matching off of GitHub repos to pretty-confidently ID anonymous coders. While set-up requires a sizable amount of compute and data, the results have gotten quite impressive. There are ways to work against this (stuff that deliberately obscures your coding style, usually by rewriting your code), but they're not that well known. And a similar thing can be done with writing style and writing samples.

Staying anonymous against high-effort attempts to discern your identity has gotten very hard, and is only likely to get harder.

At some point, all you can do is guard against the low-effort ones.

[META] Building a rationalist communication system to avoid censorship

A quiz and a day's wait before adding a new user is another option. Make it something that a regular lurker who read the rules would be able to pass easily, but a rando couldn't. SCP wiki did something like this, it seemed to help with quality control.

Rotate through 3 different quizes, or scramble the quiz order sometimes, if you want to make automated sign-ups annoying for mobs and spammers. Have the web people track the number of sign-up-quiz fails (it's a nice metric for "is there a mob at the doorstep").

(Edit: Ah, someone already proposed a more-elaborate variant using GPT-X. Simple quizes with a few mild gotcha-questions should be enough of a screen for most cases, though.)

A proposal I think I haven't seen posed is giving new members a "trial period." If an average (or randomly-selected) post doesn't have a karma score of at least X by the end of the period (or if it dips below Y at any point), they're out and their stuff is deleted. Ban them from handing out karma until after the trial, or this quickly breaks. This probably still has weird incentive consequences that I'm not seeing, though...

...it does mean having a bit of an evaporative-filter for quality-ratings, and it means links to crappy posts turn into deadlinks in just a matter of time.

Does SARS-CoV-2 utilize antibody-dependent enhancement?

I did specify long-term, which for me meant time-frames of around a year to a decade out. Honestly, I suspect you're largely right about the short-term.

Well, except I might be more optimistic about vaccination efforts. Effective vaccination pushes in the past give me some hope.

Also, the mutation rate is a good bit lower than the seasonal flu. SARS-CoV-2's point-mutations per year is around 28 substitutions, which is about 1/2 as many as the flu. Or around 1/3 the rate, at ~1.1e-3 subs per site per year, compared to flu's 3.3 subs per site per year. (Different genome lengths, hence the different answers.)

BBE W1: Personal Notetaking Desiderata Walkthrough

It's kinda weird to me how limited the options seem to be for flashcards and annotation/marginalia. While plenty of things perform the core functionality, I haven't seen anywhere near as many interesting experiments with these formats as I have with outliners.

And for flashcards particularly? You'd think it'd be the simplest damn thing to program!

I'm not much of an index card person, but there seem to be a lot of people who swear on elaborate index-card set-ups. That I haven't seen more of them implemented as software confuses me too.

(It does seem like it would be a good use of flashcards to set up a rudimentary priority queue, or stack. Possibly with added randomness. I'm honestly a little surprised it doesn't allow that.)

BBE W1: HMCM and Notetaking Systems

I think this is a good question. Here are some probable components of programmability...

  • Did it surface most of its actual functionality to users?
    • A couple different settings: Closed proprietary cloud software, API (how friendly or permissive is it?), downloadable open-source...
  • How easy (and safe!) it is to call relevant utility functions?
    • Do you need to close the software to edit it? Did they merely surface the functionality, or did they also leave functions that were highly-exposed, labeled, well-documented, and easy to use? How well do they adhere to various standards, and therefore benefit from skill-transfer? Is it easy to screw up? To revert? What's the learning curve like?
BBE W1: HMCM and Notetaking Systems


I really wish there was better flashcard and annotation/marginalia software out there! It's kinda weird to me how limited the options seem to be for both. While plenty of things perform the core functionality, I haven't seen as many interesting experiments with it as I have with, say, outliners.

While writing this post, I developed a vague suspicion that there's something in-between Annotator and Flashcard that could be pretty valuable if someone actually implemented it. This seems as good a place as any to describe it. (And if someone has already done it, or wants to do it, cool!)

Annotators and Flashcards are both often tracking an underlying dictionary-ish data-type, and a lot of flashcards seem to originate from textbooks. I have a suspicion that there should exist a good standardized-format notetaker that goes something like... this?

TextCards: 3 linked items

  • A bounded section of highlighted textbook (Any size, from a section to entire chapter. Sometimes discontinuous.)
  • An index-card laconic description (or answer)
  • A title (or question)

Sometimes, it could be used to pose standard quiz-questions (the highlighted section is just the part of the book the quiz came from, the title is the question, the description is an answer). But where it might really shine is in "Summarize Chapter X" questions; it encourages you to write along as you read the text, and if you miss something on a quiz, you can click right to the sections you were originally summarizing.

When rendered as marginalia, the small titles (until click) should make that experience more tolerable for frequent-margin-users. (Marginalia asyncing with the page seems like a really common problem, otherwise.)

For convenience, adding something that swipes all of the questions from a highlighted section of the text to form the front end of flashcards (that you then answer) seems pretty nice. For well-formatted answer-sections, you might even be able to get it to pair the two (but you'd probably need to highlight where to look). Additionally, it wouldn't be that hard for it to track which chapter's questions you're doing poorly on -and therefore what chapters you should re-read- if it knows where in the book you swiped them from. Bonus points if you can sort and cross-link notes by title, folder, tags, overlapping highlights, and/or order in text.

Presumably this is usually harder than I think it should be, because PDFs are just awful (I've dragged tables from PDFs before; I feel so sorry for Tabula!). But HTML books and ebooks don't have that problem, and often simulate a textbook-like structure.

Spiracular's Shortform Feed

At minimum it's oversimplified. That's why I called it a parable.

I appreciate the nuance in your comment even as it is. But I'm curious about the facts/narratives you have that disagree with it?

(My personal strongest murphyjitsu was that "space program defunding" was more complicated than this. If you know more about that than I do, I'd be curious to hear about it.)

Load More