Interpretation of PGE and rho in BSLMM

[Megan Ruffley]

Hello!

I was wondering if someone could share any clarification on the specific differences between PGE and rho estimated in the BSLMM, and also how to interpret them in light of trying to understand polygenicity in traits.

My current understanding is that PGE is the proportion of variance explained by major effect loci, and then rho is an approximation to PGE (this I am unclear about). I have read the rho is an approximation to PGE, but in my empirical data (several hundred traits in plant populations) often times PGE and rho are estimated to be quite different values. I cannot quite find how rho is calculated so if anyone could share or point me in the correct direction. 

Rho is a hyperparameter and I have gathered that when rho is close to 1 that indicates that the trait architecture is more appropriate for a BSVR model because a large proportion of variance is explained by the major effect loci, indicating a more monogenic architecture. When rho is close to 0, the architecture is more appropriate to be estimate in a LMM as little variance is explained by major effect loci, assuming considerable heritability. I am wondering if this is any appropriate approximation for polygeneticity?  Maybe there is something in the calculation of rho that accounts for the number of major effect loci involved?

When heritability is considerable, and PGE is low, that seems to definitely indicate a polygenic architecture. When PGE is high however, that could be due to 2 major effect loci or 155 major effect loci, and thus a high PGE does not always seems to indicate monogenicity. Again, wondering if rho estimation is considering this...

Thank you for any thoughts shared,

Megan