Heritability Estimation in TASSEL

[Mike Allwright]

Hi,

Is it possible to estimate heritability for a quantitative trait in TASSEL? The MLM_statistics file produced after an MLM analysis includes columns for the genetic variance and residual variance. Am I right in thinking that it follows that heritability can be calculated as Genetic Variance / (Genetic variance + Residual Variance)? Is it that simple?! Would this represent broad or narrow sense heritability? 

In addition, could you comment on how these genetic variance figures (and hence heritability) might be impacted by the exact model run, the nature of the kinship matrix employed and the use of optimum or no compression? 

With many thanks,

Mike Allwright. 

[David Dayan]

Hi everybody,

I've also been looking for the answer to this question.

The first row of the MLM statistics table "is for the model with no markers." I was hoping someone could clarify how the MLM estimates genetic variance without any marker data. (1) Is it an estimate of the genetic variance given the kinship matrix and phenotypes? What's going on? (2) Is it fair to use the data from this row as an estimate broad sense heritability for the trait?

- David

[Peter Bradbury]

MLM uses the kinship matrix to estimate genetic variance. Because the kinship matrix estimates an additive genetic relationship matrix, the genetic variance is actually the additive genetic variance. (Additive genetic variance) / (additive genetic variance + error variance) can be used as an estimate of heritability. See

Endelman and Jannink. 2012. G3 2:1405 - 1412 for details.

Peter

[David Dayan]

Thanks again for the quick response and the reference!

[David González]

Dear Peter,

When I ran my MLM anáisis with 300 K SNPs, I obtained a heritability of 99% (0.99999) using the ecuation (Additive genetic variance) / (additive genetic variance + error variance). Is it coherent and possible to explain?

Thanks

David