> [T]he findings suggest that people with blood type A face a 50 percent greater risk of needing oxygen support or a ventilator should they become infected with the novel coronavirus. In contrast, people with blood type O appear to have about a 50 percent reduced risk of severe COVID-19.
The prior part of the paragraph that you didn’t copy pasta was critical in understanding context here.
The study didn’t look at blood types directly, but looked at genes which express as those two blood types, finding that the genes correlate with supplemental oxygen need during severe symptoms.
In general, i regard GWAS as fishing expeditions that are more or less guaranteed to find spurious correlations. Von Neumann's old saw about "With four parameters I can fit an elephant, and with five I can make him wiggle his trunk." absolutely applies here without other information.
That said, finding associations with ACE2 and blood type regions are not totally insane!
My thoughts exactly. The article seemed to suggest that they looked at a truly ginormous number of SNP's, which suggests a high susceptibility to the xkcd "green M&M" type problem (https://xkcd.com/882/).
However, the good thing about blood type is that it should be reasonably easy/quick to do a followup where you examine a lot more patients, or even just randomly selected individuals, to see if it really holds up. As a type O person myself, I am hopeful.
as a matter of fact artificially raising the cutoff to make the results sound more convincing is a form a p-hacking
it is exactly the opposite of how p-values work. a smaller p-value does not make a result more reliable, it makes no sense whatsoever to reject a 10^-7 but trust in 10^-8
the misuse of p-values in the world is rampant - the only people that don't misuse them are those that do not use them ;-)
"The most commonly accepted threshold is p < 5 × 10−8, which is based on performing a Bonferroni correction for all the independent common SNPs across the human genome"
The reason for the low cutoff is to correct for multiple hypothesis testing.
Any what do you mean by "opposite of how p-values work"?
Obviously if P(data or more extreme|null hypothesis) = 10^-8 it's significantly less likely to see the data if the null hypothesis was true, than if P(data or more extreme| null hypothesis) = 10^-7.
My money is on the virus being an aerosol and badly ventilated indoor events accounting for most of the transmissions.
There is growing evidence for this and also that the amount of virus particles you get into your system determines how sick you get (related to your overall health).
If this is true, social distancing wouldn't help us much but disallowing gatherings of people in badly ventilated areas would.
> There is growing evidence for this and also that the amount of virus particles you get into your system determines how sick you get (related to your overall health).
Which would be a good argument for variolation.
That is, deliberately infect yourself with a very small virus dose. Small enough that you don't get seriously ill, but big enough that you get immunity.
To you and the parent, this is why I think we see large antibody counts in prison and military populations while not have as many cases. Folks were infected with a very small initial viral load and their immune systems fought it off and developed antibodies.
This is also why we see young healthy medical personnel die from the disease, 1) their immune system is degraded from harsh working conditions 2) the viral loads are many multiples higher from basically getting coughed in the face all day long
With a large enough sample size that starts to go away. We're starting to see real benefits from GWAS, it just took 1.5 decades to get the sample sizes up to a level where the data generated was worth looking at.
FWIW, The blood type correlation had already been found by a months-old study in China. But that was via ordinary blood type tests, not GWAS. I guess it lends credibility to this method that their findings are consistent with the other study.
As the proud owner of ~5 liters of type-A blood, I’m less than pleased to see this finding confirmed.
Donate blood, they'll tell you. The test is pretty simple and you can get home testing kits online if you want to do that instead; but giving blood helps other people too.
You can do a test at home using an "Eldon" kit. I did it this past weekend and it was pretty easy (other than getting the blood out). You get these special cards with reagents on that react with type A and type B blood as well as Rh+ and -, then the combination of the reactions tells you the blood type. In my case it was very obvious as the blood on the "anti B" card went very clotted whereas on the "anti A" card it was very smooth, placing me as type A.
You can donate blood, ask your doctor (insurance may not cover it), or use a simple at home test like this (Eldoncard Blood Type Test (Complete Kit) - Air Sealed Envelope, Safety Lancet, Micropipette, Cleansing Swab https://smile.amazon.com/dp/B00JFTSPMW/ref=cm_sw_r_cp_api_i_...).
I used one of the card tests and got an O- result, then when I had annual bloodwork done I asked my doctor to test it just because an O- result on the cards looks the same as a possibly defective card (card wasn’t defective, I did in fact beat the 25% odds and inherited the negative Rh factor from both my O+ parents).
It's a common "experiment" run in high school science classes. It can be done with very little equipment (just a disposable lancet to prick the skin, and a disposable card to give you the results), and lends itself to interesting lessons on genetics, antibodies, etc.
Yes there exists simple tests to figure out your bloodtype.
Where I live, they specify blood type on your passport, driving license etc. So to obtain one you need to provide proof of your bloodtype.
Had to go to a lab to get that proof.
The test itself is pretty cheap and takes 5 minutes at most. They prick your hand, take a drop of blood, and do something with that (mix it with some other chemical or something).
Wait, first and foremost age is tied to severe Covid-19
Then overall health, how many other pre-existing conditions and ongoing health issues patients already have.
Then there is the dose and the exposure the patient has.
These are by far the largest contributors.
I find it hard to believe that they have successfully disentangled all these confounding factors.
Then and only conditionally perhaps the factors that the director lists. In all, I find the blog a staggering piece of how NOT to communicate, how not to send the wrong message. Most people will read this as: "blood type A is really bad"
GWAS in general is weak stuff, even in the cited paper the odds ratios are 1.5 and 1.77 which does not indicate a particularly strong effect. Statistically may be significant, but in a practical sense not actionable. As far as GWAS studies go most turn out to be nonsense anyway.
Were any of those particular papers disproven, retracted or were not able to be reproduced?
This paper , "Genome-wide association studies: the good, the bad and the ugly" : https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4952840/ claims "the vast majority of associations identified by GWASs are extremely robust statistically and are reproducible in additional studies.".
23andMe also reported on their blog that their preliminary data also supports a protective role of blood type O:
> Individuals with O blood type are between 9-18% percent less likely than individuals with other blood types to have tested positive for COVID-19, according to the data.
and
> While it is still very early in the study, 23andMe’s preliminary investigation into genetics seems to support these findings. Comparing the research participants who reported that they tested positive for COVID-19 to those who tested negative, our researchers identified a variant in the ABO gene associated with a lower risk. (The single nucleotide polymorphism in the ABO gene is rs505922, a T at that location is associated with lower risk. The P-value for the association is 1.4e-8, OR = 0.88).
54 comments
[ 2.2 ms ] story [ 172 ms ] threadThe study didn’t look at blood types directly, but looked at genes which express as those two blood types, finding that the genes correlate with supplemental oxygen need during severe symptoms.
If so, then why would it matter whether the comment framed the issue in terms of blood type vs. those genes?
Given O is mosquitos' favorite dish, some deserved justice.
On a more serious note: the blood type studies were published months ago.
Do you have a source on that?
There is a link to the study too
That said, finding associations with ACE2 and blood type regions are not totally insane!
However, the good thing about blood type is that it should be reasonably easy/quick to do a followup where you examine a lot more patients, or even just randomly selected individuals, to see if it really holds up. As a type O person myself, I am hopeful.
as a matter of fact artificially raising the cutoff to make the results sound more convincing is a form a p-hacking
it is exactly the opposite of how p-values work. a smaller p-value does not make a result more reliable, it makes no sense whatsoever to reject a 10^-7 but trust in 10^-8
the misuse of p-values in the world is rampant - the only people that don't misuse them are those that do not use them ;-)
"The most commonly accepted threshold is p < 5 × 10−8, which is based on performing a Bonferroni correction for all the independent common SNPs across the human genome"
Any what do you mean by "opposite of how p-values work"?
Obviously if P(data or more extreme|null hypothesis) = 10^-8 it's significantly less likely to see the data if the null hypothesis was true, than if P(data or more extreme| null hypothesis) = 10^-7.
[0] https://www.nejm.org/doi/full/10.1056/NEJMoa2020283?query=fe... "Genomewide Association Study of Severe Covid-19 with Respiratory Failure"
If this is true, social distancing wouldn't help us much but disallowing gatherings of people in badly ventilated areas would.
https://www.nytimes.com/2020/06/02/opinion/coronavirus-super...
https://medicalxpress.com/news/2020-06-coronavirus-importanc...
https://youtu.be/SnI5ZsvvTrg
Which would be a good argument for variolation.
That is, deliberately infect yourself with a very small virus dose. Small enough that you don't get seriously ill, but big enough that you get immunity.
That doesn't mean it's a bad idea.
This is also why we see young healthy medical personnel die from the disease, 1) their immune system is degraded from harsh working conditions 2) the viral loads are many multiples higher from basically getting coughed in the face all day long
As the proud owner of ~5 liters of type-A blood, I’m less than pleased to see this finding confirmed.
I used one of the card tests and got an O- result, then when I had annual bloodwork done I asked my doctor to test it just because an O- result on the cards looks the same as a possibly defective card (card wasn’t defective, I did in fact beat the 25% odds and inherited the negative Rh factor from both my O+ parents).
Where I live, they specify blood type on your passport, driving license etc. So to obtain one you need to provide proof of your bloodtype.
Had to go to a lab to get that proof.
The test itself is pretty cheap and takes 5 minutes at most. They prick your hand, take a drop of blood, and do something with that (mix it with some other chemical or something).
Then overall health, how many other pre-existing conditions and ongoing health issues patients already have.
Then there is the dose and the exposure the patient has.
These are by far the largest contributors.
I find it hard to believe that they have successfully disentangled all these confounding factors.
Then and only conditionally perhaps the factors that the director lists. In all, I find the blog a staggering piece of how NOT to communicate, how not to send the wrong message. Most people will read this as: "blood type A is really bad"
GWAS in general is weak stuff, even in the cited paper the odds ratios are 1.5 and 1.77 which does not indicate a particularly strong effect. Statistically may be significant, but in a practical sense not actionable. As far as GWAS studies go most turn out to be nonsense anyway.
[citation needed]
https://massivesci.com/notes/gwas-complex-behavior-bad-biorx...
the only reason why not most GWASare listed there is because it does not pay to prove that a GWAS is a nonsense,
it is way too much work to disprove, and you don't get paid/supported for doing so
This paper , "Genome-wide association studies: the good, the bad and the ugly" : https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4952840/ claims "the vast majority of associations identified by GWASs are extremely robust statistically and are reproducible in additional studies.".
> Individuals with O blood type are between 9-18% percent less likely than individuals with other blood types to have tested positive for COVID-19, according to the data.
and
> While it is still very early in the study, 23andMe’s preliminary investigation into genetics seems to support these findings. Comparing the research participants who reported that they tested positive for COVID-19 to those who tested negative, our researchers identified a variant in the ABO gene associated with a lower risk. (The single nucleotide polymorphism in the ABO gene is rs505922, a T at that location is associated with lower risk. The P-value for the association is 1.4e-8, OR = 0.88).
See:
* https://blog.23andme.com/23andme-research/23andme-finds-evid... * https://www.snpedia.com/index.php/Rs505922
A p-value of 1.4e-8 is significant enough when there is a prior hypothesis. And the odds-ratio looks normal too.