1000 Genomes Low Coverage Data Inside A Gwas Study

Question

[Sander Timmer]

For many of the HapMap CEU cell lines there is low coverage genotypes available since the pilot of the 1000 genomes. Though still in most cases people would prefer to use the well studied and high confidence tag-SNPs from HapMap instead of these new 1000 genomes SNPs.

What are the main reasons for doing so? I can understand computational time is a reason and also genotype quality is a reason I could think off. What should men expect when using the 1000 genomes SNPs instead of the HapMap SNPs? And what are the caveats that people should consider before doing this? Are there any additional steps required to be taken in terms of increasing the confidence of the given genotypes?

I'm not planning on doing anything like this myself but I would like to read about this a bit more. So any general comments of pointers to must-read papers about this are highly appreciated!

Answer 1

[lh3]

One of the key goals of the 1000g project is to produce a denser reference panel for GWAS. I have seen several high-profile GWAS papers using the 1000g data. And from a few talks, my impression is having 1000g data (at least a more recent version of ~600 samples) is preferred. If you want to see more theoretical analyses, have a look at Goncalo's paper published in Genome Research. There may be better papers.

Answer 2

[Larry_Parnell]

One issue with many of the SNPs that the 1000G project uncovered is their minor allele frequency. If the frequency is too low, one loses power to detect an association with the phenotype of interest. To counter this, there is the theory that a rarer allele will likely have greater (phenotypic) effect, thus power may not suffer so much. Also, it is possible that the rarer allele is tagged by a more common one, but whereas a large proportion of carriers of the rare minor allele may show disease risk, far fewer of the carriers of the minor allele of the more common tag SNP will show that disease risk.