|
Post by gfarries on Jun 3, 2015 9:26:16 GMT
I have been using GCTA to calculate the heritability of several traits in horses using both the additive GRM and dominant GRM.
For some traits I am finding almost zero additive heritability, but using the additive+dominant GRM getting heritabilities of 0.3+.
Source Variance SE V(G1) 0.000032 3.800158 V(G2) 12.214193 11.414265 V(e) 20.880480 11.182725 Vp 33.094705 2.921379 V(G1)/Vp 0.000001 0.114827 V(G2)/Vp 0.369068 0.333898
Sum of V(G)/Vp 0.369069 0.351796
I am concerned that the standard error for the Vg is very high. My n=289, however my sample population would have very low genetic diversity, Ne for this breed of horse has been estimated at 100.
Any feedback would be greatly appreciated as I am quite new to GCTA.
|
|
|
Post by Zhihong Zhu on Jun 3, 2015 13:14:21 GMT
Hi there,
Because the SE is very large, the SNP heritability, additive or dominance, is not significant. Did you identify significant markers when you did GWAS? What was the genetic inflation (lambda_GC)? If p values of GWAS are not inflated, it may be because 1) genetic variation at causal variants is small 2) sample size is small.
Cheers, Zhihong
|
|
|
Post by gfarries on Jun 8, 2015 11:19:38 GMT
Thank you zhihong, that is very helpful.
Yes there were a few significant GWAS hits but only when examining the P2df (genotype as opposed to allelic associations). I will check the lambda estimate for those associations.
I wish to perform a power calculation to determine the sample size required. To do so I need to estimate the var_pi as I suspect mine will be quite different to the default 2e-5 for human studies. I realise this is a very basic question, but could you help direct me to how to estimate var_pi from my GRM? Is there a tutorial or explanation available?
|
|
|
Post by Zhihong Zhu on Jun 9, 2015 7:50:42 GMT
|
|