Welderufael, B.G.1,2, De Koning, D.J.1, Fikse, W.F.1, Strandberg, E.1, Franzén, J.3 and Christensen, O.F.2, 1Swedish University of Agricultural Sciences, Department of Animal Breeding and Genetics, Box 231, 750 07 Uppsala, Sweden, 2Aarhus University , Department of Molecular Biology and Genetics, Building K23, 8830 Tjele, Denmark, 3Stockholm University , Department of Statistics, 106 91 Stockholm , 106 91 Stockholm , Sweden.
Genetic evaluation of mastitis is performed either with cross-sectional or longitudinal models. In this study we aim to develop better longitudinal models using simulated SCC (Somatic Cell Count) which usually is used as a proxy to label clinical mastitis. Data was simulated for mastitis liability and recovery for two scenarios (28% and 95% mastitis cases/lactation) and two daughter groups of 60 and 150 per sire in 1200 herds. Weekly observations for SCC were simulated assuming a baseline curve for non-mastitic cows and deviations in case of a mastitis event. Binary data was created to define presence or absence of mastitis as 1 if the simulated SCC was above pre-specified boundary and 0 otherwise. The boundary was allowed to vary along the lactation curve modeled by a spline function with a multiple of 10 or 15. The dynamic nature of the SCC was taken in to consideration with the longitudinal approach and the patterns were captured by modelling transition probability of moving across the boundary. Thus, a transition from below to above the boundary is an indicator of the probability to contract mastitis, and a transition from above to below the boundary is an indicator of the recovery process. Estimated breeding values for mastitis liabilities and recovery were calculated in DMU.
Our preliminary results showed the correlation between true and estimated breeding value (accuracy) for the simulated mastitis liability was 0.72 which is as good as the estimations based on clinical mastitis. Though the estimation accuracy for recovery (0.42) was not as high as for mastitis liability, the transition probability model enables us to generate breeding values for mastitis recovery process.