849
An approach to genomic analysis of longitudinal data using random regression
Genetic evlauations of 305-days milk yield has been more accurately estimated using random regression models (RRM). We propose the use of random regression coefficients as phenotype for genomic evaluation of longitudinal milk production data. Pedigree based estimated breeding values of 1) milk yield at 305 day (P305), 2) three independent random regression coefficient (Coef305) and 3) cummulative breeding values estimated with RRM from day 6 to 305 (RRM305) are deregressed and used as pseudo-phenotypes. GEBV's (gP305, gCoef305 and gRRM305) are estimated with a genomic-polygenic model. Pair comparison using spearman rank correlatons of GEBV between pseudo-phenotypes and 10 fold crossvalidation were used to estimate predictive ability. Spearman rank correlation were 0.85 between gP305 and both gCoef305 and gRRM305; and 0.94 between gRRM305 and gCoef305. Predictive ability was 0.74 and 0.63 for gRRM305 and gCoef305. Deregressed random regression coefficient can be used in genomic evaluations.
Keywords:
genomic evaluation; lactation curve; SNP
