Using regression trees to highlight the synergistic and antagonistic effects of farming practices on intrinsic quality indicators of cow’s milk
Abstract
The intrinsic quality of milk is a multifactorial and multidimensional concept that is influenced by various farming practices. These latter can interact with each other to have synergistic or antagonistic effects on the different milk compounds or properties (quality indicators). This observational study was set out to identify and prioritise some farming practices that influence milk quality indicators, and to highlight the practices that have synergistic or antagonistic effects on these indicators. The study used two datasets with 233 bulk tank milks from 113 French dairy farms, both containing data on their intrinsic quality (33 quality indicators measured) and the farming practices applied on the farms at the time of collection, such as herd characteristics, feeding, housing, and milking management, and milk storage conditions. We then used regression trees to predict milk quality indicators based on combinations of the farming practices. Among the 33 milk quality indicators, the regression trees modelled 9 from combinations of farming practices with an R2 ≥ 0.30. Those regression trees enabled the identification of the most influential farming practices, the validation of practices previously reported in controlled experimental trials under commercial farm conditions, and the characterisation of synergistic and antagonistic interactions among practices. The dominant breed and the main forage in the diet were the most influential factors, followed by the proportion of concentrates, the lactation stage, and the proportion of primiparous cows in the herd. Housing conditions and milking practices did not emerge as influential on the wellperforming trees. Certain practices, such as pasture grazing or the choice of Montbéliarde breed, appeared to have an overall positive effect on the milk quality indicators they influenced. However, due to antagonistic effects, it is challenging to identify practices that enhance all indicators simultaneously. Therefore, trade-offs will need to be made based on the farmers’ goals and the targeted dairy product, as different products require distinct milk qualities. Moreover, for many indicators, the part of the variability not explained by the regression trees remained significant, and the trees could be refined from larger databases and from practices surveyed specifically for each indicator.
Related
- Phenotypic plasticity segregates antagonistic species but undermines their coexistence
- Revisiting the Endophytic Fungi Residing Within Axonopus compressus: Molecular Identification, Antagonistic Potential, and Extracellular Enzyme Activities
- Clodronate and trehalose exhibit a paradoxical antagonistic interaction that suppresses proliferation of canine mammary tumour cells
- Fungistatic and non-fungistatic soils as reservoirs of bacteria antagonistic to Fusarium graminearum
- Environment-dependent genetic architecture of seed vigor reveals antagonistic allelic effects under water deficit in rapeseed (Brassica napus)
BibTeX
@article{reycadilhac2026using,
title = {{Using regression trees to highlight the synergistic and antagonistic effects of farming practices on intrinsic quality indicators of cow’s milk}},
author = {Lucille Rey-Cadilhac and Claire Laurent and M. Gelé and Stéphanie Léger and Anne Ferlay},
journal = {Animal},
year = {2026},
month = sep,
doi = {10.1016/j.animal.2026.101947},
url = {https://hal.inrae.fr/hal-05770642v1},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - JOUR
TI - Using regression trees to highlight the synergistic and antagonistic effects of farming practices on intrinsic quality indicators of cow’s milk
AU - Lucille Rey-Cadilhac
AU - Claire Laurent
AU - M. Gelé
AU - Stéphanie Léger
AU - Anne Ferlay
PY - 2026
DA - 2026/09/30
JO - Animal
DO - 10.1016/j.animal.2026.101947
UR - https://hal.inrae.fr/hal-05770642v1
AB - The intrinsic quality of milk is a multifactorial and multidimensional concept that is influenced by various farming practices. These latter can interact with each other to have synergistic or antagonistic effects on the different milk compounds or properties (quality indicators). This observational study was set out to identify and prioritise some farming practices that influence milk quality indicators, and to highlight the practices that have synergistic or antagonistic effects on these indicators. The study used two datasets with 233 bulk tank milks from 113 French dairy farms, both containing data on their intrinsic quality (33 quality indicators measured) and the farming practices applied on the farms at the time of collection, such as herd characteristics, feeding, housing, and milking management, and milk storage conditions. We then used regression trees to predict milk quality indicators based on combinations of the farming practices. Among the 33 milk quality indicators, the regression trees modelled 9 from combinations of farming practices with an R2 ≥ 0.30. Those regression trees enabled the identification of the most influential farming practices, the validation of practices previously reported in controlled experimental trials under commercial farm conditions, and the characterisation of synergistic and antagonistic interactions among practices. The dominant breed and the main forage in the diet were the most influential factors, followed by the proportion of concentrates, the lactation stage, and the proportion of primiparous cows in the herd. Housing conditions and milking practices did not emerge as influential on the wellperforming trees. Certain practices, such as pasture grazing or the choice of Montbéliarde breed, appeared to have an overall positive effect on the milk quality indicators they influenced. However, due to antagonistic effects, it is challenging to identify practices that enhance all indicators simultaneously. Therefore, trade-offs will need to be made based on the farmers’ goals and the targeted dairy product, as different products require distinct milk qualities. Moreover, for many indicators, the part of the variability not explained by the regression trees remained significant, and the trees could be refined from larger databases and from practices surveyed specifically for each indicator.
SN - 1751-7311
SN - 1751-732X
ER -