Strategies to combine novel traits across countries: example of heat stress

Abstract

Nowadays, novel traits are of great interest. However, phenotypes are siloed and mainly not shared. Heat stress is becoming problematic affecting animals’ performances and their well-being. Heat stress tolerance as a novel trait is only addressed by isolated within-country research studies. Integration and combination of local and foreign information sources is needed for better accuracy genetic evaluations. Therefore, this study was aimed to test the potential combination of sources of external information towards the evaluation of heat stress tolerance of dairy cattle. Long-term cow performances linked to environmental descriptors (weather parameters as proxy to climate change) collected over 10 years under the temperate conditions of the Walloon Region of Belgium and the hotter and warm Mediterranean conditions of Andalusia and Castile-La-Mancha Spanish regions were available. A total of 1,604,775 milk, fat, and protein test-day (TD) records linked to average daily temperature humidity (THI) values for 3-day lag before each TD were considered. Under a first strategy considering free-access to raw-data (phenotype and pedigree), a joint evaluation was firstly run using reaction norm models where production traits were considered as function of THI. A Belgian and a Spanish evaluation were also run using the same model. An alternative strategy considering only access to external information (i.e. regression coefficients for additive genetic effects (â and their associated REL)) was tested. In this case, foreign â and their REL resulting from the Spanish evaluation were first converted to the Belgian trait and thereafter integrated in the Belgian evaluation using a Bayesian approach. Rank correlations between regression coefficients, â (of the 1,104 bulls having daughters only in Spain) estimated by Belgian evaluation and â estimated by the joint evaluation were moderate (<=0.70). Corresponding rank correlations between â estimated by joint and Bayesian evaluations were significantly higher (ranging from 0.967 to 0.998), indicating that the Bayesian evaluation integrating external information was in good concordance with the joint evaluation. Results from this study indicated that the integration of external information via the Bayesian approach has a good potential to improve the genetic evaluation of sparse and siloed novel traits.