GWAS with few samples

Question

[LNA]

Hello,

I have SNP data of a Illumina HumanOmni5-4 BeadChip (4.2 million SNPs) for 28 samples with a quantitative trait. I have no experience doing GWAS studies and I need some advice. So far, I performed the quality control steps suggested by the GWASTools Vignette.

Which testing procedure is best for this kind of data? It seems, such a small number of samples leads to no significant p-values. Any way to improve this? Should I use the quantitative trait directly with linear regression based testing or does it make any sense to group the samples and do a case-control study?

Any suggestion would be of great help to me!

Thanks, Lna

[WouterDeCoster]

Unless your effect size is extremely huge, 28 samples will never be sufficient to get significant/meaningful and accurate association results.

[LNA]

I have to analyze that data. Since increasing the sample size is not an option, I want to know what is the best I can do?

Answer 1

[coleman_jonathan]

Assuming your QT is a genuine continuous variable, you should use the QT. It has greater power (read this: http://www.ncbi.nlm.nih.gov/pubmed/11987478)