Parameters and covariance between breeding values in ancestral regression.

GARCÍA-BACCINO, C. A.; MUNILLA, S.; FORNERIS, N.S.; CANTET, R. J. C.; ROGBERG-MUÑOZ, A.

Auckland

Congreso; 11 World Congress of Genetics applied to Livestock Production.; 2018

WCGALP New Zealnd Society for Animal Production

The ancestral regression (AR, Cantet et al., 2017) is a quantitative genetics model thatprovides a unified framework of the contributions from pedigree and genomic information for predicting BV under non-independent causal variants, under a Gaussian causal (Pearl, 2000) distribution for BV. Thus, AR represents the inheritance of BV by a recursive system of structural equations such as in the animal model (Henderson, 1984). The parameters of AR are path coefficients that are identifiable with the aid of a dense panel of SNP or sequences. A distinctive advantage of the ancestral regression is that the inverse of the covariance structure is computed in a linear fashion as by a simple extension of the rules of Henderson (1984), such that direct inversion is not required. As in the regular animal model, the information to predict BV under AR is by covariances between relatives, which in turn are functions of pedigree and genomic identifiable parameters. This expository paper discusses the meaning of the parameters in AR, and the covariance between the BV of related animals. Certain developments in AR require a Gaussian distribution for BV. For reasons of space, the involved proof of asymptotic normality of BV will be presented elsewhere.