genherit1.asreml.R - Gmat

[lucasph97]

Hi there Sam and Ari,
I have been looking at your code on heritability as we are aiming to get these outcomes as well for a breeding program. I was wondering if you could clarify for me if the Gmat corresponds to the numerator/genomic relationship matrix used in the vm() function in the asreml syntax?
Say for instance the genetic model is random =~ diag(Env):vm(Geno, GRM) + diag(Env):ide(Geno), would then the Gmat in genherit1.asreml.R be GRM?

Many thanks in advance
regards,
Lucas

[ariavda]

Hi Lucas,

Firstly, genherit1 does not handle multi-site trials. I have a version in development, genherit, that would allow multi-site analysis like the model you present. The Gmat argument is for compound terms like yours (or a single term for which GRM is Gmat for example, and where narrow sense heritability is rquired)), but without the diag(Env). That is
vm(Geno, GRM) + ide(Geno) so that we want the broad sense heritability (getting the correct PEV of course will require the only argument to be specified. The code doesn't work out the full variance matrix. It has to be done manually outside genherit1.
lastly, you can do what you want by passing arguments to predict via ..., for example levels to go through levels of Env. Given your expertise, I think you can work out how to "loop" through Env to get the heritabilities you want. Gmat will need to be calculated for each level of Env.

I hope this helps,

Ari

[lucasph97]

Hi Ari,
Many thanks for your reply.
Thanks for the clarification on genherit1 working only for single site trials. I had noticed that but was trying to figure out the logics on the single-site approach to move to the multi-site case. Actually I have been playing around with loops across Env's to get the PEV (using the only argument as you mentioned) so it seems like I'm on the right track
Still have to work around on getting the Gmat for each Env, but with the feedback that you have provided I think I can make my way through
Thanks again for taking the time to reply. Looking forward for the multi-site version of genherit!

[ariavda]

Hi Lucas, Thanks for your email. I hope it is all going smoothly. There are issues regarding heritability with pedigrees because the lines present at each site or Env will be different in general. But the pedigree enables prediction of all effects. So should the heritability calculation use all the lines or just the lines that are actually in the trial? I think Valeria showed the latter is what you should do some time ago. She can comment as I have copied her into this email. The same issue needs to be thought about if a GRM is used. Regards, Ari Prof. Ari Verbyla AV Data Analytics Email: ***@***.*** From: Lucas Peitton ***@***.***> Sent: Thursday, 14 November 2024 12:17 AM To: biometryhub/lmmtools ***@***.***> Cc: Ari Verbyla ***@***.***>; Comment ***@***.***> Subject: Re: [biometryhub/lmmtools] genherit1.asreml.R - Gmat (Issue #1) Hi Ari, Many thanks for your reply. Thanks for the clarification on genherit1 working only for single site trials. I had noticed that but was trying to figure out the logics on the single-site approach to move to the multi-site case. Actually I have been playing around with loops across Env's to get the PEV (using the only argument as you mentioned) so it seems like I'm on the right track Still have to work around on getting the Gmat for each Env, but with the feedback that you have provided I think I can make my way through Thanks again for taking the time to reply. Looking forward for the multi-site version of genherit! — Reply to this email directly, view it on GitHub<#1 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/BM4SRAPSV2KSP72KYGMTG632ANJ33AVCNFSM6AAAAABRUCYMAKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINZTGY3DONZWGY>. You are receiving this because you commented.Message ID: ***@***.******@***.***>>