5 ms·

I work in this field. The reasons for these problems are probably relatively simple. 1) Current GWASs focus on common variants, but the population explosion o

I work in this field. The reasons for these problems are probably relatively simple. 1) Current GWASs focus on common variants, but the population explosion over the last few thousand years means it's likely that most human genetic diversity is rare. 2) Current GWASs focus on the marginal effects of variation at single-nucleotide sites, but it's likely that there are complex interactions between different variations. These misplaced foci of current GWASs have more to do with what we could do with current technology than with what we ought to do, given these biological realities: sequencing was so expensive that it was hard to determine the rare variants; sample sizes were so small that we had to use simplifying statistical assumptions like independent phenotypic effects from each nucleotide variation. Those technological limitations are changing, now. The "thoughtful geneticists" the article refers to aren't just wringing their hands about this. They are working on both of these issues. (I'm one of them.)

randallsquared 17y ago
Interestingly, the article mentions both of your listed problems as possible escapes from the conclusion.
bioweek 17y ago
That sounds fascinating. Are you hiring ...
carbocation 17y ago
I also work in this field. This article is interesting and gets some things right (although is overly alarmist, in my opinion). There are opinions in this article that I disagree with, however. For example, I disagree with the statement that "... classical Mendelian genetics based on family studies has identified far more disease-risk genes with larger effects than GWAS research has so far." First, under selection, one would not expect common variants to have tremendous, deleterious effects in general. Second, it is not true that Mendelian studies are superior to GWAS in, say, lipoprotein genetics. There, virtually all of the genes known from studies of Mendelian disorders have been re-discovered through GWAS, with many new findings that were previously unknown. Furthermore, a substantial portion of lipid heritability (~20%) could be explained by GWAS-identified loci as of January 2009. There there is little reason to believe that common variants (>=5% minor allele frequency) should explain all of heritability; so far, so good. The upcoming 100,000 individual lipids GWAS should do a pretty good job of sucking at the marrow for the remaining common variants. We work with the tools that we have. SNP-chips enabled GWAS. Now that "next-gen" sequencing (worst buzzword ever) has brought down costs, the thousand genomes project will provide us with a list of nearly all variants with MAF>1%. Well-funded groups can afford to deep sequence their population of choice. These provide information that is complementary to the information obtained by GWAS. In conclusion, GWAS is good for finding common variants. Imputation from thousand genomes should be good for finding low-frequency (>1% MAF) variants. And sequencing is good for finding rare variants. There is no crisis.
narag 17y ago
I'm very confused by the article. Not sure if it's because the language or because it's not very clear. On the one hand it says that there are too little evidence of "hereditability" to be useful to cure diseases. On the other hand it says that there is a lot of evidence, to the point of being "disturbing", of racial traits. What am I reading wrong?
- carbocation 17y ago
  Regarding the missing heritability, they are claiming that this is a deficiency of GWAS. It then goes on to talk about sequencing, which it says will find lots of evidence for miscegenation, etc. (To which I say, sure, but what of it?) In case you are not familiar with the methods, the difference between GWAS/SNP analysis and sequencing is this: GWAS relies on surveying ~500,000 (up to 6 million now, but it doesn't really matter) known single nucleotide polymorphisms. So you might ask, "which nucleotides of this subset of the human genome differ in a systematic way between cases and controls." With sequencing, you are surveying all nucleotides, not just a subset of 500,000. So you instead ask "which nucleotides out of the 3 billion (haploid) differ between cases and controls?"
  - narag 17y ago
    ...will find lots of evidence for miscegenation, etc. (To which I say, sure, but what of it?) Thank you for the explanation. I'd love to know more about my DNA, specially if it says that I have a lot of far relatives around the world, I can't understand that's a problem.

randallsquared 17y ago

Interestingly, the article mentions both of your listed problems as possible escapes from the conclusion.

bioweek 17y ago

That sounds fascinating. Are you hiring ...

carbocation 17y ago

I also work in this field. This article is interesting and gets some things right (although is overly alarmist, in my opinion). There are opinions in this article that I disagree with, however. For example, I disagree with the statement that "... classical Mendelian genetics based on family studies has identified far more disease-risk genes with larger effects than GWAS research has so far." First, under selection, one would not expect common variants to have tremendous, deleterious effects in general. Second, it is not true that Mendelian studies are superior to GWAS in, say, lipoprotein genetics. There, virtually all of the genes known from studies of Mendelian disorders have been re-discovered through GWAS, with many new findings that were previously unknown. Furthermore, a substantial portion of lipid heritability (~20%) could be explained by GWAS-identified loci as of January 2009. There there is little reason to believe that common variants (>=5% minor allele frequency) should explain all of heritability; so far, so good. The upcoming 100,000 individual lipids GWAS should do a pretty good job of sucking at the marrow for the remaining common variants. We work with the tools that we have. SNP-chips enabled GWAS. Now that "next-gen" sequencing (worst buzzword ever) has brought down costs, the thousand genomes project will provide us with a list of nearly all variants with MAF>1%. Well-funded groups can afford to deep sequence their population of choice. These provide information that is complementary to the information obtained by GWAS. In conclusion, GWAS is good for finding common variants. Imputation from thousand genomes should be good for finding low-frequency (>1% MAF) variants. And sequencing is good for finding rare variants. There is no crisis.

narag 17y ago

I'm very confused by the article. Not sure if it's because the language or because it's not very clear. On the one hand it says that there are too little evidence of "hereditability" to be useful to cure diseases. On the other hand it says that there is a lot of evidence, to the point of being "disturbing", of racial traits. What am I reading wrong?

carbocation 17y ago
Regarding the missing heritability, they are claiming that this is a deficiency of GWAS. It then goes on to talk about sequencing, which it says will find lots of evidence for miscegenation, etc. (To which I say, sure, but what of it?) In case you are not familiar with the methods, the difference between GWAS/SNP analysis and sequencing is this: GWAS relies on surveying ~500,000 (up to 6 million now, but it doesn't really matter) known single nucleotide polymorphisms. So you might ask, "which nucleotides of this subset of the human genome differ in a systematic way between cases and controls." With sequencing, you are surveying all nucleotides, not just a subset of 500,000. So you instead ask "which nucleotides out of the 3 billion (haploid) differ between cases and controls?"
- narag 17y ago
  ...will find lots of evidence for miscegenation, etc. (To which I say, sure, but what of it?) Thank you for the explanation. I'd love to know more about my DNA, specially if it says that I have a lot of far relatives around the world, I can't understand that's a problem.