|
Post by roberto on Dec 30, 2015 10:56:31 GMT
Hello,
A paper in PNAS by Krishna Kumar et al (http://www.pnas.org/content/early/2015/12/17/1520109113.abstract) suggests that all estimations of h2 obtained with GCTA should be re-evaluated. They also have a proposition to 'denoise' GCTA's GRMs. Could you comment on this paper? (there is probably a detailed response in preparation?)
thank you very much in advance,
roberto
|
|
|
Post by Jian Yang on Jan 19, 2016 23:25:01 GMT
The “denoising” approach will lead to underestimation of the variance explained by SNPs. For instance, it is shown in their Figure 6 that the heritability estimate is unbiased without adjustment (Figure 6A; an estimate of 0.62 is not significantly different from the true parameter 0.65 given SE = 0.22) and biased when the GRM is adjusted using the denoising approach (Figure 6B; an estimate of 0.17 is significantly different from 0.65 given SE = 0.22). Please see gcta.freeforums.net/thread/257/commentary-krishna-kumar-pnas-2015 for a full commentary of the Krishna Kumar et al. paper.
|
|
|
Post by roberto on Jan 20, 2016 16:03:39 GMT
Thank you! reading the paper right now.
|
|