Conditional Analysis Of Snps (Ex From Gwas)

Question

[Thomas]

Hi all,

Within complex diseases like T2D and obesity, several genome-wide hits has been established as true associations. I know that there have been several attempts to locate the functional variation by fine-mapping the surrounding regions by genotyping Hapmap SNPs (+/-200kb) or sequencing of the locus.

Analysis of such data normally ends up with conditional analysis by making association study of the novel SNP and then condition on the lead SNP.

My Qs is: How do we establish whether we have found a true independent hit? or how do we establish that a novel SNP explains the association seen for the lead SNP?

We have done a lot of these conditional analysis and several times we see that surrounding interesting SNPs (ex a missense) gets more significant when conditioning on the lead SNP. Ex p-values going from 10-3 to 10-4. But is this just coincidence? can we make some power calculations? or what to do.

My concern is that these kind of studies gets very subjective when conclusion are made.

OK... that was a lot of Qs. In general I would like to hear if anyone has experience with conditional analysis of SNPs and how they interpret these results... Or even better; if you have another way of approaching this area.

Thanks for any kind of contribution!

All the best Thomas

Answer 1

[David Quigley]

When an association has a very modest effect on heritable susceptibility of a complex trait, it is extremely difficult to establish that a candidate variation is the true source of a signal. The editors of Nature Genetics said as much in an editorial accompanying one of their recent GWAS-packed issues. Quoting briefly:

In complex traits, demonstrating causation may be impossible, but there should at least be evidence that the variants identified in a GWAS show a consistent correlation with gene expression or gene regulation that is also highly correlated with the human trait originally assayed. At this point, other systems can be used experimentally to provide functional support: for example, mapping mammalian quantitative trait loci in the syntenic region, investigating deletions of the syntenic region, natural variants and mouse knockouts of nearby genes, fish or cell knockdown of a nearby gene and in vitro expression studies with constructs of nearby genes.

I think the essential point is that although fine-mapping can get you a compelling candidate, to "establish that a novel SNP explains the association seen for the lead SNP" is going to require a coordinated program of further experimental work rather than ever-greater amounts of statistical analysis. There's no getting away from the need for hypothesis-driven experimentation in this situation.

Answer 2

[Larry_Parnell]

I agree with David's advice. A good study will reach a point of transition from statistical and population genetics to cell biology and molecular investigation of the biological role(s) of the alleles. This often necessitates cloning sequence containing the alleles - perhaps even haplotypes - into appropriate vectors and expression in appropriate cell types. The vast majority of such work involves allele-specific analysis of expression and function and ignores haplotype effects. Haplotypes though should not summarily be dismissed but for some quite practical reasons might not be explored.

[Thomas]

OK... Thanks to both of you for sharing your thoughts. I think I see now where this is heading. thanks Thomas

[Larry_Parnell]

You're welcome. In terms of an independent hit, as you ask, you can also consider replication of the genetic association results in another population.

[Thomas]

Yes... also something that we are considering. Our results are based on sequencing with follow up genotyping. We are thinking of asking collaborators to impute 1000G into there dataset and compare the association pattern with ours.