Epsilon Archive for Student Projects

Abstract

Genotype by environment interaction for somatic cell count and mastitis in the first lactation of the Swedish Holstein breed was studied, using a reaction norm model and multiple trait analysis. Data from the Swedish milk-recording scheme containing more than 200 000 cows having their first calving from 1995-01-01 and onwards was used. Somatic cell count was defined as the average value of the monthly milk sampling results until 150 days after calving expressed in 10 000 cells per millilitre and transformed to a logarithmic scale to the base 10 (LSCC). Mastitis was defined as an all-or-none trait observed from 10 days before calving to 150 days after calving. Environments were defined as: herd-year averages of 305-day protein yield, somatic cell count, and mastitis, all measured in first lactation; and herd size, expressed as the number of cows with first calving per year. Furthermore, overall herd size expressed as the average of 1995-2000 herd-year sizes was used. The multiple trait analysis was done with two models using the highest and lowest quartiles of the environments herd-year protein yield, herd-year somatic cell count, herd-year mastitis, herd size as well as overall herd size. The genetic correlation for LSCC and mastitis between low and high environments indicated GxE for LSCC in somatic cell count environment. Variances of the slope and the level of the reaction norm were analyzed by regressing phenotypic values of somatic cell count and masti-tis on herd-year environments of protein yield, somatic cell count, mastitis and on the envi-ronment overall herd size. Significant genetic variation in slope was also detected for LSCC in somatic cell count environments and the correlation between predicted offspring perform-ance in low and high somatic cell count environments showed GxE and indicated re-ranking of sires. The heritability of somatic cell count and mastitis estimated as functions of the envi-ronments tended to be lowest in average environments and increased with the distance from the average. Neither the multiple trait analysis nor the reaction norm model provided us with complete results when using the environmental scale mastitis. This is probably due to that the mastitis frequency is close to zero in low mastitis environments, resulting in a lack of pheno-typic variation.