Gwa Case Control Studies In Related Individuals

Question

[Sudeep]

Are there any established methods of case control studies in related individuals ? From doing some googling I found about EMMAX, but the approach is developed for quantitative traits than binary triats, and can be adopted for use binary traits. My question is are there are there any specific approaches developed for GWAS on binary traits in related samples ? I cannot remove related samples because if I do that I will finally end up with very few samples

[Sudeep]

update.. I found that generally case - control studies are not necessarily done on related individuals because of lose of power (for test statistic), instead related individuals are usually used for quantitative traits

Answer 1

[David Quigley]

The most common way to perform genetic analysis of related individuals is a linkage study. In this case, you explicity take advantage of the pedigrees of your affected/unaffected subjects to map loci associated with the condition. You could look into the Transmission Disequilibrium Test (see papers by Spielman) for details, though this is a very large and old subject in genetics so there are lots of places you could begin for an overview. If you're definitely interested in family-based association studies, I suggest this review by Benyamin, Visscher and McRae in Pharmacogenomics or papers by Laird and Lange such as this review in Nature Reviews: Genetics or this one in Advances in Genetics. They have methodological recommendations and discusse some specific packages designed for family-based association studies.

Google is great, but in general, to get guidance on software and methods when entering a new area I recommend doing some Pubmed searches to find recently published reviews.

[Sudeep]

Thanks a lot, the second and third papers were helpful (did not have full text access for the first one although the abstract looked interesting )

Answer 2

[Genotepes]

Hi

check out Mary-Sarah McPeek web page - University of Chicago. They have developped software for related data - they "just" (a lot of fantastic work behind this just) explicitely model the variance-covariance matrix based on relatedness and kinship.

Actually, the TDT based tests are good too but quite restrictive if you don't think there is much stratification in your sample. It sort of answers to a slightly different questtion. You may combine both, especially if you don't have much founders typed.

Genotepes