I'm assuming "time since vaccine second dose" in their chart should read "time since last vaccine dose"?
It's going to be interesting to see if/how genetics can predict who is susceptible to (symptomatic) COVID. Might depend on the variant, too: I know someone whose spouse got COVID twice, but they only got infected after a lunch meeting with a friend... For those who are susceptible (i.e. most of us), the question isn't so much "will doing x ensure that I never get COVID", but whether doing x will a. delay getting infected (this was more important early on when health outcomes were worse), b. reduce the severity of an infection (outcomes appeared to be worse for people who were exposed to larger quantities of virus), and c. reduce the frequency of getting infected.
Actually, I suspect that chart hasn't been updated in a while and indeed it means "since second dose".
In addition to your points, there's also the possibility that a SARS-CoV-2 infection today is protective against some other, even scarier infection in the future. There's speculation that the 1918 Flu was particularly serious in people who hadn't been born (and infected) by an earlier pandemic.
I would like to respectfully point out an error you made in the interpretation of this paper.
You mention "According to their preliminary data, people with HLA-DQB1*06 alleles were less likely to experience PCR-confirmed breakthrough infection. 23andme reports HLA-DQB1 status here. For me, it’s all Ts and Gs (no As). “
In the paper they do say “This AA variant” but that is not alleles AA, but amino acid.
HLA alleles were imputed from genotype data and the two alleles best tagging HLADQ B*6 were rs9271374 Glu71Arg (position 32661752, Chr 6, build 38) and rs1130456 DQB1-125A/S (Position 32659602, chr 6, Build 38). Neither of these SNPs appear to be in my raw data file of 23andme, version 3. Maybe they are in other versions. In essence you need to look at your specific results for those 2 SNPs (which can be triallelic…either A, C or T). HLA-DQB1*06 has a glycine at position 125, whereas other alleles common in the genotyped population possess either alanine (HLA-DQB1*02 and *04 alleles) or serine (HLA-DQB1*05). Thus, this amino acid variant (rs1130456 with Gly) is synonymous with the presence of HLA-DQB1*06 in our dataset.
These two rsid#s are not present (searching by position or rsid#). Sorry, once I found out these not in my raw data, I did not take the time to sleuth out what allele corresponds to Gly, Ala or Ser. Long story short: you cannot determine your HLA DQB1 status from 23andme, v3.
From the paper
These are SNPs located within 10 kilobases (kb) of HLA-DQ genes (Fig. 2) and in linkage disequilibrium within our multi-ancestry cohort (r 2 = 0.65). The distribution of P values (Extended Data Fig. 3a) and beta coefficients (Extended Data Fig. 3b) for all genotyped and imputed variants across this locus show a clear correlation in genetic architecture between these two antibody responses (Spearman’s rho coefficient 0.90 and 0.93 for P values and beta coefficients, respectively) correlated through linkage disequilibrium (measured through r 2 ).
HLA imputation was performed using the Multi-Ethnic HLA reference panel (version 1.0 2021) available on the Michigan Imputation Server46 using recommended settings. Phasing of multi-allelic HLA alleles was undertaken using PHASE (version 2.1.1)47. HLA typing was also performed using PCR sequence-specific primers (SSPs) at the Histogenetic laboratory, Oxford University NHS Foundation Trust.
This amino acid AA variant (DQB1-125A/S) denotes the presence of either an alanine or a serine at position 125 of the HLA-DQB1 protein according to international ImMunoGeneTics (IMGT) project coordinates. The index-associated variant from the primary analysis (rs1130456) was equally associated with the anti-RBD titers in this analysis (beta = 0.27 and s.e. = 0.04). Other variants imputed using the specific HLA imputation algorithm were identified as being marginally more significantly associated than rs1130456, with the new lead being rs9273817 (P = 2.4 × 10−9, beta = 0.27 and s.e. = 0.04).
You are right! Thank you, thank you! I am always on the lookout for ways to see how my own data corresponds to something in a scientific paper, but it appears this time it's not so simple. I've updated the post to point to your excellent comment.
I'm assuming "time since vaccine second dose" in their chart should read "time since last vaccine dose"?
It's going to be interesting to see if/how genetics can predict who is susceptible to (symptomatic) COVID. Might depend on the variant, too: I know someone whose spouse got COVID twice, but they only got infected after a lunch meeting with a friend... For those who are susceptible (i.e. most of us), the question isn't so much "will doing x ensure that I never get COVID", but whether doing x will a. delay getting infected (this was more important early on when health outcomes were worse), b. reduce the severity of an infection (outcomes appeared to be worse for people who were exposed to larger quantities of virus), and c. reduce the frequency of getting infected.
Actually, I suspect that chart hasn't been updated in a while and indeed it means "since second dose".
In addition to your points, there's also the possibility that a SARS-CoV-2 infection today is protective against some other, even scarier infection in the future. There's speculation that the 1918 Flu was particularly serious in people who hadn't been born (and infected) by an earlier pandemic.
Re: Long COVID, a friend wrote a lengthy (scary) discussion here that I'll probably comment on later: https://substack.com/notes/post/p-125665724
I would like to respectfully point out an error you made in the interpretation of this paper.
You mention "According to their preliminary data, people with HLA-DQB1*06 alleles were less likely to experience PCR-confirmed breakthrough infection. 23andme reports HLA-DQB1 status here. For me, it’s all Ts and Gs (no As). “
In the paper they do say “This AA variant” but that is not alleles AA, but amino acid.
HLA alleles were imputed from genotype data and the two alleles best tagging HLADQ B*6 were rs9271374 Glu71Arg (position 32661752, Chr 6, build 38) and rs1130456 DQB1-125A/S (Position 32659602, chr 6, Build 38). Neither of these SNPs appear to be in my raw data file of 23andme, version 3. Maybe they are in other versions. In essence you need to look at your specific results for those 2 SNPs (which can be triallelic…either A, C or T). HLA-DQB1*06 has a glycine at position 125, whereas other alleles common in the genotyped population possess either alanine (HLA-DQB1*02 and *04 alleles) or serine (HLA-DQB1*05). Thus, this amino acid variant (rs1130456 with Gly) is synonymous with the presence of HLA-DQB1*06 in our dataset.
These two rsid#s are not present (searching by position or rsid#). Sorry, once I found out these not in my raw data, I did not take the time to sleuth out what allele corresponds to Gly, Ala or Ser. Long story short: you cannot determine your HLA DQB1 status from 23andme, v3.
From the paper
These are SNPs located within 10 kilobases (kb) of HLA-DQ genes (Fig. 2) and in linkage disequilibrium within our multi-ancestry cohort (r 2 = 0.65). The distribution of P values (Extended Data Fig. 3a) and beta coefficients (Extended Data Fig. 3b) for all genotyped and imputed variants across this locus show a clear correlation in genetic architecture between these two antibody responses (Spearman’s rho coefficient 0.90 and 0.93 for P values and beta coefficients, respectively) correlated through linkage disequilibrium (measured through r 2 ).
HLA imputation was performed using the Multi-Ethnic HLA reference panel (version 1.0 2021) available on the Michigan Imputation Server46 using recommended settings. Phasing of multi-allelic HLA alleles was undertaken using PHASE (version 2.1.1)47. HLA typing was also performed using PCR sequence-specific primers (SSPs) at the Histogenetic laboratory, Oxford University NHS Foundation Trust.
This amino acid AA variant (DQB1-125A/S) denotes the presence of either an alanine or a serine at position 125 of the HLA-DQB1 protein according to international ImMunoGeneTics (IMGT) project coordinates. The index-associated variant from the primary analysis (rs1130456) was equally associated with the anti-RBD titers in this analysis (beta = 0.27 and s.e. = 0.04). Other variants imputed using the specific HLA imputation algorithm were identified as being marginally more significantly associated than rs1130456, with the new lead being rs9273817 (P = 2.4 × 10−9, beta = 0.27 and s.e. = 0.04).
You are right! Thank you, thank you! I am always on the lookout for ways to see how my own data corresponds to something in a scientific paper, but it appears this time it's not so simple. I've updated the post to point to your excellent comment.