|
Post by kelsey on Apr 26, 2018 1:53:28 GMT
I am interested in whether the SNPs used to create GRMs can be different to the SNPs used in MLMA-LOCO.
I have 170,000 candidate regulatory variants, and I'm going to test whether they have any other phenotypic effects. These variants cover 2,500 1mb regions throughout the genome, and I am going to add a background of ~30k SNPs from a chip set, so 200,000 total. I have 110,000 animal samples and an additional 19 million imputed SNP.
For the GRMs, should I use these 200,000 variants or should I subset it to only include say the 30k chip set SNP, given the particular distribution of the 170,000 regulatory variants.
I'm interested in any discussion on these approaches.
Thank you.
|
|
|
Post by zhilizheng on May 4, 2018 4:04:48 GMT
I would like to have all variants given well QC condition. If the variants are selected randomly enough, the GRM will be very close.
However, we can perform the checking steps, whether the GRM differ a lot or not.
|
|