Comparison of country-specific predictions of feed intake and methane emissions in sheep using different proxies

Abstract

Ruminants are often singled out as being the main culprits when it comes to greenhouse gas (GHG) emissions, for methane (CH₄) in particular. However, with their diets based on forage and grazing, ruminants have a role to play to limit the feed-food competition. Sheep breeders are open to the prospect of including both feed efficiency and GHG emissions in their breeding programmes and whether or not it is for the purpose of genetic (or genomic) selection, the acquisition of new phenotypes for feed efficiency and GHG emissions are essential. Currently, devices recording GHG emissions and individual feed intake of animals reared indoors remains too expensive for most sheep breeders worldwide. In this study, research groups from six countries (UK (Scotland), France, Norway, Ireland, New Zealand and Uruguay) gathered their results obtained in different breeds to identify the most promising proxy measurements of feed intake and methane emissions. Despite the fact that each group set up their own protocol, there were several points in common: most feed intake trials were performed during 6 weeks on growing animals, and GHG emissions were all recorded with portable accumulation chambers (PACs). Different traits, in addition to feed intake and GHG emissions, were recorded and considered as putative proxies (body composition, growth, bodyweight, feeding behaviour, body condition score), as well as sheep genotypes and ruminal microbiota. Models' goodness of fit were estimated on training sets, whereas their prediction accuracy was assessed on actual validation datasets. The comparison of training and validation accuracies obtained with each dataset highlighted the well-documented problem of overfitting, particularly with microbiota data. In general, validation prediction accuracies were higher for feed intake than for the two feed efficiency criteria (residual feed intake and feed conversion ratio) investigated. The best predictions for feed intake were obtained when body weight and the average number of feeding events per day were included in the models (R²_valid=0.78). Methane emissions were predicted with the highest accuracy when feed intake was considered among the proxies. Prediction accuracies for methane emissions obtained with the metagenome were higher than with sheep genome, although this accuracy remains quite low (R_valid=0.32).

By comparing independent results from six countries, we highlighted that the recording of body weights, feeding behaviour and fixed effects are crucial for the prediction of feed intake. However, feed efficiency and methane emissions are relatively difficult to predict well, with the proxy measurements assessed here.