Using family data in Mixed Linear Model Association Analyses (MLMA)

Question

[coleman_jonathan]

Background: Using multiple genomic relatedness matrices (GRMs; one for the whole cohort, and one just capturing familial relationships), it is possible to capture SNP-heritability in family data (as detailed on the GCTA forums and in Zaitlen et al, 2013: http://gcta.freeforums.net/thread/241/gcta-greml-analysis-family-data http://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1003520)

Question: How can this be incorporated into an MLMA analysis?

Potential solutions: Naively, I can see three options, but I am unsure which is the best:

1. Using the GRM for the whole cohort (seems vulnerable to confounding from shared environmental factors)
2. Fitting both GRMs in the model (straightforward, but not convinced this appropriately captures the gross shared genomic relationship)
3. Performing some kind of subtraction of the familial GRM from the whole cohort GRM (most likely through generating breeding values from each GRM first), and fitting the result as a random effect capturing the gross genetic relationship (akin to the described method for estimating heritability, but not straightforward without performing the MLMA outside of existing software).

If anyone has any experience with this or insight into the correct approach, I would be very grateful to hear it!

Answer 1

[coleman_jonathan]

For anyone interested in the answer to this question, look at Jian Yang's response here: http://gcta.freeforums.net/thread/323/family-data-mlma