PVE more than 100%

36 views

Skip to first unread message

Muhammad Habib Widyawan

unread,

Aug 14, 2024, 1:54:00 PM8/14/24

to R/qtl discussion

Hi Karl,

I mapped two QTLs for a single trait on different chromosomes using CIM, the broad sense heritability of the trait was around 90%. However, when using CIM and fitqtl function independently, the sum of two PVE of those QTLs are more than 100%, is that possible to overestimate the PVE?

Thanks!

Karl Broman

unread,

Aug 14, 2024, 4:01:21 PM8/14/24

to R/qtl discussion

The PVE estimates provided by fitqtl is just a transformation of the LOD score, and only account for a single QTL at a time.

I wouldn't put much trust in the estimates, and certainly the total percent variance explained can't be more than 100%.

As stated in the help file for fitqtl:

The estimated percent variances explained for the QTL are simply transformations of the conditional LOD scores by the formula h2=1−10^[−(2/n)LOD]. While these may be reasonable for unlinked, additive QTL, they can be completely wrong in the case of linked QTL, but we don't currently have any alternative.