Reducing Gwas Snp List To ~Independent Tag Snp List

Question

[Occam]

I am conducting a genomewide methylation QTL study. This entails 2M (methylation markers) * 650K (SNPs) tests, which is intractable on our cluster. Thus, we want to reduce the number of SNPs using only a subsample of approximately independent (tag) SNPs. The Tagger server and Haploview crash when I upload the Plink format GWAS data, so I wonder if perhaps Tagger is not well-suited for operations including so many SNPs? Is this the case, or am I making an error otherwise? If Tagger is not appropriate for this task, could you recommend another tool?

[Occam]

this is a dead thread, but for posterity - i eventually discovered pruning functions in plink ("--indep" followed by "--extract") which sped computation time ~30-fold compared to kluging through a perl scripted looped (by chromosome) haploview call.

Answer 1

[Gustavo]

Perhaps MAGMA could be of use?

Answer 2

[Larry_Parnell]

You could order your GWAS SNPs by chromosome and upload one chromosome at a time.

It is important to use a program for Tag-SNP calculations that is recognized as good and that uses a robust dataset for those calculations. In other words, you want to be certain that the right population data is behind the LD/Tag-SNP calculations, or one that is capable of using your own GWAS data to make those calculations.

From our colleagues, working with Affy 6.0 genotype data + ~1.6 imputed SNPs on ~850 individuals:

To calculate LD in among (the study's) SNPs, we uploaded the data to Haploview (pedigree file, phenotype, and genotype). Haploview then calculated the LD among SNPs using the maximal set of unrelated individuals. Some of the plots that were generated showed that sometimes this works for the imputed SNPs and other times it doesn’t, so (the person with whom I corresponded) is not sure when imputed SNPs are included, but in the regions in which there was interest it looked like more often than not the imputed SNPs had LD calculated.

[Occam]

Thanks for the input. Have made some headway using Haploview doing each Chr separately.

Answer 3

[Fayue1015]

Hi, have you solved with this problem? I would recommend our tool "fasttagger" for a try, it is much efficient than current popular tools