The prediction of genetic values based on genomic and phenotypic information plays an important role in dairy cattle breeding. This thesis investigates if an improved prediction can be obtained when in addition the metabolome is considered. Data for the three system-levels were (a) simulated using a systems biology approach and (b) experimentally collected (~1300 cows). An integrative bioinformatics approach for data analysis was developed. Concluding, the metabolome provides a deeper insight into the relationships between different levels, whose exploitation can lead to improved prediction.