Estimate signifiant loci in GWAS

Question

[alessandrotestori7]

Hi! Assume that a Genome Wide Association Study (GWAS) yields three significant SNPs related to a disease; another GWAS yields two more SNPs (different from the previous ones) related to the same disease and another GWAS finally yields four significant SNPs involved in the same disease, but these SNPs were already found in the previous studies all but one. This scenario suggests that the more GWAS datasets are generated, the more significant associated SNPs will be discovered. Is there a way to compute how many new SNPs will be discovered in the next generation of n GWAS datasets? How many significant SNPs are yet to be discovered thanks to GWAS? Are there softwares which make the calculation? Please help me with some suggestions. Thanks.

Answer 1

[WouterDeCoster]

With increasing sample size the power of your analysis will increase, as such identifying more SNPs passing genomewide multiple testing correction. However, it's hard to estimate what the "expected but a priori unknown genetic effect" is, in terms of effect size and population frequency vs heterogeneity. If you google for "power calculation gwas" you'll find many hits pointing you to resources which might be of interest for you.

[alessandrotestori7]

Thanks! Power, however, is related to single SNPs, am I right?

[WouterDeCoster]

Power will dependent on your population size, the allelic heterogeneity, the clinical heterogeneity and the effect size of the expected variants.

[stolarek.ir]

not necessarily power relates to single SNPs. You might observe some effect if 2 independent (not in LD) snps are present. Google LAMPLINK software, I think it's one of the most recent approaches for dealing with multiple SNPs in GWAS and lack of power for using SNPs combinations

Answer 2

[alessandrotestori7]

I recently found "INPower" package in Bioconductor which seems to implement this!