Given the fact SARS had a similar profile, this should be a reasonable assumption.
I’m not saying it’s at the same standard as being peer reviewed, though it’s more likely than not that the conclusions from this data will remain true based on both the data and the correlation.
> I’d be happy to change the title, though it is basically quoted from the article conclusion.
For future reference, Hacker News guidelines call for using the article's original title as the title of the submission. https://news.ycombinator.com/newsguidelines.html (the moderators seem to have fixed this for you)
I would like a hypothesis tested on individual alleles as I outlined in another comment: it's not that weak correlation if looked at that way, especially since eg. A* groups are present in Caucasian population in ~45% compared to ~35% in Asian population (numbers pulled from another site cited in that other comment of mine).
If you're referring to the odds ratio between group A control and group B covid, then that's not a Bayes factor AFAIK. A Bayes factor is to compare two hypothesis. You can have an odds ratio of 1.00001 which is very significant (eg. female/male birth ratio)
No, they didn't look at severity. They did report ABO distributions for 206 deaths in Wuhan, which could be considered a proxy for severity but likely has a lot of confounding variables, such age, wealth, etc. Those data were in-line with the top-line results, but that's a small sample.
They are looking at data from hospitals, so I'd expect mild cases to be excluded.
Semantic nuances seem to be very relevant to the applicability (or not) of this selection to the exact wording of the result: is a SARS-CoV-2 infection that remains harmless because it is overcome before spreading from throat to lungs classified as a light case of Covid-19 or is that label reserved for more serious forms of infection?
If ABO type is independent from age/wealth/etc., then those variables don't need to be controlled (although I guess that should indeed be investigated/confirmed first).
By the way (and not flamebaiting here) this is one of the reasons why "this is natural selection" is the phrase that makes me the most angry in all of this. Death by flu is selection for random shit that cause you to be better off against only one thing and that is: that exact desease causing strain.
The worst thing about Covid 19 to me is how it selects against old people, and even worse: Friendly, sociable old people.
Indeed. Those sort of naturalistic[0] and appeal to nature[1] fallacies are infuriating. One advantage of being human is that "this is how things are" doesn't have to mean "this is how things must be."
Nature also optimizes for far more variables than humans do. We min-max for a handful of KPIs at most. Nature optimizes for everything that an organism needs to survive, including in once-in-a-lifetime circumstances. It must do so since most organisms don't have dedicated support infrastructures. Cars don't need to concern themselves with reproducing, a jaguar does.
Sure, there are genuine inefficiencies like the laryngeal nerve but there also are plenty of apparently useless things that turn out to have some niche importance such as the appendix.
It's definitely possible. That's what large companies are for; they subsume some part of the economy into themselves, intelligently design it in ways that could never be found through an unplanned search, and are greatly rewarded if they do it well. Amazon, for example, found one of the best possible designs for both direct-to-consumer shipping and cloud systems.
But we don't usually think about it that way. Most people who advocate for a planned economy don't want that kind of planning; they think it's bad to innovate in things like how jobs are structured, bad have a single large company in charge of most of your shipping, and bad that Jeff Bezos can build such a good company that his stake in it is worth billions. Planned economy advocates that I see generally want the economy to work how they specifically would prefer, without much regard for whether other people think that's the best way.
In theory you have bankruptcy, limited liability and the corporate person as devices meant to insulate human beings from the harsh natural environment that makes businesses work. Unfortunately, businesses can't stand being treated harshly (and politicians don't have the guts to do it) so we have bailouts, protective regulation, and federal troubled asset purchases, kind of defeating the point.
True!
If it would, then male baldness would have been "natural de-selected" a long time ago. But since men are getting bald mostly after they had their offspring, partners (nature) cannot select on this...
I'd argue that killing off a source of wisdom and knowledge that hasn't been fully passed down hurts the species as a whole. This isn't a counter to your point, just another perspective.
At least selecting against religious people who congregate
Seems that e.g., Buddhists regularly practicing month-long meditation retreats or Christians practicing 40-day contemplations might see a protective effect selecting them in, tho such practices may reduce reproduction rates, so what is the net balance?
Conversely, the countries in the EU that have been affected the worse, are heavily religious countries that encourage regular mass gatherings as part of their faith.
There are second order effects like: Old people taking care of the grand children so the parents can do other things. Old people sharing their experiences directly or indirectly with the younger generation etc.
This is why lazy Darwinian hand-waving is - let's be blunt - stupid and naive.
By the time you get to a complex human culture, everything is about second and third order effects.
We're here debating whether or not it's Darwinian that Covid-19 disproportionately affects old people, while the second and third order effects are a recession (if we're lucky and clever) and a depression (if we're neither).
Given where we are as a species right now, both of those will have a much bigger impact on long-term human viability than any immediate deaths.
> The worst thing about Covid 19 to me is how it selects against old people, and even worse: Friendly, sociable old people.
That's cherry-picking IMO. It also selects against people with chronic diseases, high blood pressure, unhealthy habits like smoking and against people with poor hygiene.
Your dismissive "of course" is not warranted for the sentence that says "The worst thing about Covid 19 to me is how it selects against old people".
Two ways for Covid 19 to not have that characteristic:
1) It doesn't select against anyone, meaning it doesn't kill people.
2) It doesn't select against old people, meaning it's less selective, meaning (keeping the same mortality rate) it kills less old and more younger people.
Is there some particular reason why this would not pass peer review in some low level journal? Are the statistics off? Or is the data fabricated? Because it will be published and it will pass the peer review if it satisfies those basic conditions. Peer review just means that someone else read the paper and judged whether the content seems reasonable. I don't think it is productive to call an article bogus because it is not peer reviewed. The judgement should be based on content.
I am a little rusty on my probability basics now but I was trying to remember how one would go about answering the following question regarding Wuhan numbers cited in this study:
"Given a person has type A blood type, what is the probability of that person having covid19"
I remember this being something like P(E1 intersection E2)/P(E2) where E2 = event that a person has type A blood and E1 = event that the person has covid19.
In case of Wuhan 32.16% of the population has type A blood. So E2=0.3216. To get E1 we would need total infections/total population for the city. We don't have that number but is this thinking correct?
In particular, as you describe it, you'd then need to solve for P(E1 ∩ E2) by extrapolating from the existing result of P(E2 | E1) present in the article, then plug it into that formula you've provided.
(Bayes' law streamlines the whole process by combining the two-step calculation into one.)
edit: a slight nitpick: E2 is an event ("this person has type A blood"); P(E2) is a number.
This has no more value than being a "fun fact" (sorry for the word 'fun' in the context of a pandemic and death numbers)..
a) it is something you don't have any control
b) the diff (odds, bayes factor or any similar definition) is not that great to have an effect on public policy (something like, let's focus our resources on region A, since region B has less risk kind of change)
So, let's all enjoy this study as food for our curiosity, and move on with all the usual measures of social distancing, extra hygiene control and all.. Especially dangerous, if even one crazy person goes out because he has a certain blood type and helps the disease spread
I recommend reading the review below on known effects of blood groups on infection. Not super surprising in that blood group may have general effects. I can't think how this would work mechanistically without knowing the glycans on ACE2/TRMPSS2 or the spike, and if they carry antigen (I would expect not).
Assuming the equal distribution of homo- and heterozygotes, it looks like the numbers (A/AB +20%, B +3%, 0 -33%) suggest the following:
• 00 (homozygote 0, but that's the only way someone is a 0) is at -33% "chance" of catching SARS-Cov-2
• B0 (heterozygote B) is at -10%
• BB (homozygote B) is at +5%
• AB (well, heterozygote, obviously) is at +20%
• A0 (heterozygote A) is at <17%
• AA (homozygote A) is at ~33%
Basically, the hypothesis would be that presence of an A allele contributes significantly to increased risk (~17%), B is mostly neutral (+1-2%), a 0 allele contributes to decreased risk (-17%).
Considering the distribution in healthy individuals (I haven't run the numbers but only used roughly 30% for each group except AB which is at 10% — I've only done the math in my head so I am way more off than that, and it would be quite unlikely for the effect to be so linear as my breakdown above suggests) from the study, everything roughly lines up.
It would be interesting if a study could confirm that, but they'd need parents blood types for every individual to be able to get that.
I've seen mention of open data sets that include research articles, but is there any data set to look at? It would be quite interesting since the distribution of blood types supposedly differs strongly among "ethnicities" if https://www.livescience.com/36559-common-blood-type-donation... is to be trusted — their "Asian" number for AB is also at 7% compared to 9% in their "health population" number.
If there was data just documenting whatever findings there are for patients, it would help in independent "researchers" find correlations that might not be obvious to others.
Who knows. Maybe this is a fluke maybe not. When you notice a correlation you study it. Sometimes you discover it isn't significant and stop further study. Sometimes it is significant so you study more.
That isn't to say your question is bad, just that it could be years before we understand.
I mean blood type is literally molecules exposed on the outside of blood cells. I believe they are exposed on other cells, not just on blood cells. Viruses bind to receptors on the cell wall in order to infiltrate the cell. Perhaps the A blood type receptor makes it easier for the virus to enter?
Blood type antigens are complex carbohydrates, and A/B/O is the difference between N-acetyl-Galactosamine/Galactose/nothing at the end of a sugar chain.
So let's talk about Coronavirus and sugars. SARS-COV-2 (and the prequel for that matter) are relatively unique for viruses because they don't (as far as we know) use sugars (sialic acids to be specific) to bind to cells, unlike all of our other favourite viruses (e.g. influenza).
BUT that's not the only place it could matter. Maybe sugars help organise the membrane to increase multivalent reactions, and there's some galectin that binds to B antigen.
Or the virus itself somehow carries the antigen, and A/B protects against furin processing?
Lots of possibilities. First thing I would check for is co-expression of ABO with ACE2. If it's not in the same cells, then I have NO idea how it works.
Well.. I don't know what blood type I am but it is somehow comforting to know that there are at least some people out there who have a better chance at not getting sick.
Slight nitpick: I find the use of 'normal people' in the article a bit distracting. I appreciate the fact that the authors are not native English speakers but using 'the general population' was more appropriate here. However, being a blood type 'O', I am not sure whether this consoles me.
81 comments
[ 4.1 ms ] story [ 139 ms ] thread3694 Control (Wuhan area)
A: 1188 (32.16%)
B: 920 (24.90%)
AB: 336 (9.10%)
O: 1250 (33.84%)
COVID-19 1888 patients
A: 715 (37.87%)
B: 494 (26.17%)
AB: 193 (10.22%)
O: 486 (25.74%)
The difference may be related to triaging criteria. E.g. different ethnic groups overrepresented in given hospital.
Not peer reviewed, but an interesting observation nevertheless. There will be more data soon.
I’m not saying it’s at the same standard as being peer reviewed, though it’s more likely than not that the conclusions from this data will remain true based on both the data and the correlation.
For future reference, Hacker News guidelines call for using the article's original title as the title of the submission. https://news.ycombinator.com/newsguidelines.html (the moderators seem to have fixed this for you)
edit: added wikipedia link
"ABO group in 3694 normal people in Wuhan showed a distribution of 32.16%, 24.90%, 9.10% and 33.84% for A, B, AB and O, respectively"
- Reported odds ratios[0] of 1.2 for A and 0.67 for O.
- This is a preprint, so it has not been peer reviewed.
- They compared blood type of patients with that of the general population in the region.
- Data is from 2,173 patients at three hospitals. The data varied between them, but Wuhan had the overwhelming majority (1,775, or 81%) of cases.
- Since these are all patients, they presumably are weighted toward the more extreme responses. There were minimal age and gender differences, though.
- A similar response for type O was apparently reported for SARS.
0: https://en.wikipedia.org/wiki/Odds_ratio
Semantic nuances seem to be very relevant to the applicability (or not) of this selection to the exact wording of the result: is a SARS-CoV-2 infection that remains harmless because it is overcome before spreading from throat to lungs classified as a light case of Covid-19 or is that label reserved for more serious forms of infection?
The worst thing about Covid 19 to me is how it selects against old people, and even worse: Friendly, sociable old people.
0: https://en.wikipedia.org/wiki/Naturalistic_fallacy
1: https://en.wikipedia.org/wiki/Appeal_to_nature
There has never been an economy like the latter, and it's a tragic failure of human imagination to believe that it isn't possible.
Sure, there are genuine inefficiencies like the laryngeal nerve but there also are plenty of apparently useless things that turn out to have some niche importance such as the appendix.
But we don't usually think about it that way. Most people who advocate for a planned economy don't want that kind of planning; they think it's bad to innovate in things like how jobs are structured, bad have a single large company in charge of most of your shipping, and bad that Jeff Bezos can build such a good company that his stake in it is worth billions. Planned economy advocates that I see generally want the economy to work how they specifically would prefer, without much regard for whether other people think that's the best way.
There are some kin effects if elderly are killed, but natural selection otherwise doesn’t apply much for the elderly.
balding men on the other hand, not confidence inspiring.
Seems that e.g., Buddhists regularly practicing month-long meditation retreats or Christians practicing 40-day contemplations might see a protective effect selecting them in, tho such practices may reduce reproduction rates, so what is the net balance?
Not judging this, just pointing it out.
Eg. Italy is at 72%, Spain at 49%, USA at 69%, France at 30% on https://en.wikipedia.org/wiki/Importance_of_religion_by_coun...
I would consider them countries where social gatherings are the norm, and I always related that to climate, rather than religion.
But that is, in fact, natural selection, isn't it?
Natural selection ≠ selection for traits that benefit society.
But still unproven then, because it's a hypothesis and not a theory...
Natural selection may be “natural”. But nature is unbelievably cruel.
By the time you get to a complex human culture, everything is about second and third order effects.
We're here debating whether or not it's Darwinian that Covid-19 disproportionately affects old people, while the second and third order effects are a recession (if we're lucky and clever) and a depression (if we're neither).
Given where we are as a species right now, both of those will have a much bigger impact on long-term human viability than any immediate deaths.
That's cherry-picking IMO. It also selects against people with chronic diseases, high blood pressure, unhealthy habits like smoking and against people with poor hygiene.
Otherwise fine.
You think Covid 19 would be better if it selected against young people?
For whatever it's worth, if Covid-19 killed more young people, it might have been taken more seriously earlier.
Two ways for Covid 19 to not have that characteristic:
1) It doesn't select against anyone, meaning it doesn't kill people.
2) It doesn't select against old people, meaning it's less selective, meaning (keeping the same mortality rate) it kills less old and more younger people.
https://en.wikipedia.org/wiki/Blood_type_distribution_by_cou...
There is absolutely 0 peer reviewed papers or double blind controlled studies that prove this.
Do you have a basis for saying its bogus? This is a new result.
How would you even do a double blind control study. Expose people to the virus?
"Given a person has type A blood type, what is the probability of that person having covid19"
I remember this being something like P(E1 intersection E2)/P(E2) where E2 = event that a person has type A blood and E1 = event that the person has covid19.
In case of Wuhan 32.16% of the population has type A blood. So E2=0.3216. To get E1 we would need total infections/total population for the city. We don't have that number but is this thinking correct?
In particular, as you describe it, you'd then need to solve for P(E1 ∩ E2) by extrapolating from the existing result of P(E2 | E1) present in the article, then plug it into that formula you've provided.
(Bayes' law streamlines the whole process by combining the two-step calculation into one.)
edit: a slight nitpick: E2 is an event ("this person has type A blood"); P(E2) is a number.
So, let's all enjoy this study as food for our curiosity, and move on with all the usual measures of social distancing, extra hygiene control and all.. Especially dangerous, if even one crazy person goes out because he has a certain blood type and helps the disease spread
Imagine you are debugging a random connection loss issue in prod. You notice that the errors are much more frequent on server-05.
Ok, fun fact. So what?
What is special about that server? Does it serve customers from a different geographic zone? Is it in another network location?
https://www.ncbi.nlm.nih.gov/m/pubmed/26085552
• 00 (homozygote 0, but that's the only way someone is a 0) is at -33% "chance" of catching SARS-Cov-2
• B0 (heterozygote B) is at -10%
• BB (homozygote B) is at +5%
• AB (well, heterozygote, obviously) is at +20%
• A0 (heterozygote A) is at <17%
• AA (homozygote A) is at ~33%
Basically, the hypothesis would be that presence of an A allele contributes significantly to increased risk (~17%), B is mostly neutral (+1-2%), a 0 allele contributes to decreased risk (-17%).
Considering the distribution in healthy individuals (I haven't run the numbers but only used roughly 30% for each group except AB which is at 10% — I've only done the math in my head so I am way more off than that, and it would be quite unlikely for the effect to be so linear as my breakdown above suggests) from the study, everything roughly lines up.
It would be interesting if a study could confirm that, but they'd need parents blood types for every individual to be able to get that.
I've seen mention of open data sets that include research articles, but is there any data set to look at? It would be quite interesting since the distribution of blood types supposedly differs strongly among "ethnicities" if https://www.livescience.com/36559-common-blood-type-donation... is to be trusted — their "Asian" number for AB is also at 7% compared to 9% in their "health population" number.
If there was data just documenting whatever findings there are for patients, it would help in independent "researchers" find correlations that might not be obvious to others.
That isn't to say your question is bad, just that it could be years before we understand.
Here's the answer: We have no idea!
Blood type antigens are complex carbohydrates, and A/B/O is the difference between N-acetyl-Galactosamine/Galactose/nothing at the end of a sugar chain.
So let's talk about Coronavirus and sugars. SARS-COV-2 (and the prequel for that matter) are relatively unique for viruses because they don't (as far as we know) use sugars (sialic acids to be specific) to bind to cells, unlike all of our other favourite viruses (e.g. influenza).
BUT that's not the only place it could matter. Maybe sugars help organise the membrane to increase multivalent reactions, and there's some galectin that binds to B antigen.
Or the virus itself somehow carries the antigen, and A/B protects against furin processing?
Lots of possibilities. First thing I would check for is co-expression of ABO with ACE2. If it's not in the same cells, then I have NO idea how it works.