What Kinds Of Hypotheses Can Be Supported By Significant Snps In A Gwas Study

Question

[kumar.vinod81]

Significant SNPs through GWAS are fine mapped type or we need some further work to prove the actual causative variation for the trait under study. I've done a GWAS in a plant species and identified a set of 30 SNPs showing significant signals for a susceptibility index. It means that the top most SNP is the actual cause of the trait variation or yet it need some further work to proof the actual region underlying the trait. What other we can propose with the significant SNPs? Thanks,

[Pierre Lindenbaum]

This question was answered and then closed . I reopened it.

[kumar.vinod81]

Thank you very much....in future I'll take care of it..

[brentp]

"we propose to go fishing. hypothesis: we will catch something"

Answer 1

[Jeremy Leipzig]

My primitive understanding from my brief stint in agriculture is that most hits seen in plant GWAS are, in order of least common to most common:

• the causative SNP
• in LD with the causative SNP, you're close to it and have picked up the haplotype
• spurious associations related to population structure effects (e.g. fungal resistance is more common in certain plant subpopulations with hairy leaves so the hairy leaf SNP is a top hit on the manhattan plot). Correcting for population effects is something of an art.

The next step I saw was positional cloning, but this was usually done in arabidopsis.

[kumar.vinod81]

Thanks for answer. Can you please little more elaborate your answer, I am getting it but not in complete way?

[cmdcolin]

Here's a blog post about some of these topics...shows a couple techniques that can be used to 'bootstrap' based on the population distributions of SNPs (similar to the third recommendation by Jeremy here). http://www.cureffi.org/2013/01/29/sampling-a-matching-distribution-for-bootstrapping/

I imagine that the second problem might be solved by QTL type techniques but i am not sure exactly how

[kumar.vinod81]

Thanks a lot for valuable information. The first information is really interesting and important.

Answer 2

[Giovanni M Dall'Olio]

The SNPs identified in a GWAS are not the real causal variant of the phenotype, but they are rather SNPs within a LD block with the real causal variant. So, sorry but you still have some work to do to discover what is causing your phenotype :-)

A few years ago, in this forum, there it has been an effort to contribute to a collaborative paper on methods for Post-GWAS functional characterization of risk loci. We started adding a section about computational methods to do such analysis. We added a general overview of the methods available, and a table of the tools that were available at the moment. Unfortunately, in the end the contributions from the online community were not taken into account :-( , and the paper was published in Nature Genetics without them. You may refer to the published paper for a reference on the functional characterization of the SNPs you have identified, and can also look at the wikigene draft for some references on the computational approaches you can use now. Anyway, if you want to look for more recent literature, you should search for the keywords Post-GWAS characterization, SNP prioritization, and candidate gene approach.

[kumar.vinod81]

Thank you very much. This is really I am searching from last few days, and now I am thinking that I will be able to do something with my GWAS results. Thanks Giovanni

Answer 3

[swbarnes2]

In order to really prove that a SNP is causative; that is, that the variant is necessary and sufficient to cause the variation observed, you need benchwork. You can't prove such a thing computationally.

[kumar.vinod81]

Yes, you are absolutely right, but before my question on this blog, I was unaware in finding out the direction that where to go from here? But now I found something to go from here. Thanks a lot for your answer....