|
Post by hnorpois on Apr 16, 2014 12:15:38 GMT
Hello,
I have a question concerning the random effect g in the MLM. In the model y=Xb+g+e (with y: nx1 vector of phenotypes b: fixed effects g: random effect and e as error term) I dont understand how to measure g, that seems to rely on the GRM (a n x n matrix for n phenotypes). g is a n x 1 matrix. Is g the first component of a pca (on the n x n of GRM) or the first component of a multidimensional scaling? I dont think so. So how do you get a "polygenetic effect" for each id (corresponding to a phenotype)?
I am sorry to ask this question here but I did not find an answer in the papers. Thanks Hermann
|
|
|
Post by Jian Yang on Apr 16, 2014 12:35:13 GMT
For one individual, g = sum(w*u) with w being the standardised SNP genotype code and u being the SNP effect. g can be regarded as the total genetic effect of an individual. Or, some people would call it the polygenetic effect. g can be estimated using BLUP given that you know the var(g). Please see Yang et al. 2011 AJHG (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3014363/) for more details.
|
|
|
Post by hnorpois on Apr 16, 2014 16:57:52 GMT
This was very helpful. Thanks. But... I still have a problem. It is mentioned that "u is a vector of snp effects". Is this the same as the breeding value?
|
|
|
Post by Jian Yang on Apr 17, 2014 6:18:44 GMT
"breeding value" = g = sum(w*u)
|
|
|
Post by hnorpois on Apr 17, 2014 9:14:34 GMT
Thank you but I still have a u-problem. How is it measured? Thanks.
|
|
|
Post by Jian Yang on Apr 17, 2014 11:42:55 GMT
u is the SNP effect. u is not measured. u can be estimated using BLUP given var(g).
|
|