The paper “Machine learning modelling for weighting and ranking of multiple variables related to genotype-environment interaction: innovative protocol proposal for selecting breeding cows in wetlands“, published in The Journal of Agricultural Science, has been chosen as the latest Editorial Highlight and is freely available.

Traversing a time marked by frequent revisionist intentions, the revaluation of findings, and the high speed of information suggested by modern algorithms, we recognise the real need to incorporate different approaches into the strong bioenvironmental system, shaped by the fundamental interaction between cattle–environment–humans, with branches leading towards strategic aspects of today, such as: the production of animal-based protein, the rational use of sensitive environmental systems, animal welfare, traditional socio-economic sectors, alongside the powerful tools of artificial intelligence, inferential statistics, and mathematical equations.

Our theories focus on the marked link that exists in the genotype–environment interaction, particularly among bovine biotypes raised in areas that pose significant challenges to modern livestock farming, such as freshwater wetlands and semi-arid regions. In our geography, these areas represent vast rural expanses where the main challenge is to produce calves with adequate efficiency, while at the same time ensuring animal welfare and environmental harmony. In Paraguay, the cattle breeds used in genetic improvement, selection, and commercialisation programmes are imported, which may not permanently meet all the aforementioned challenges. Therefore, the focus and premise of our studies is to thoroughly employ bioenvironmental data engineering along with modern computational tools, to develop a novel protocol for selecting breeding cows, based on bovine populations traditionally raised in freshwater wetlands, including the set of imported breeds which, according to commercial enterprises, are those that show “better adaptation” to the warm, humid, forage-based environment of the region. All of this is to identify potential individuals with superior adaptive capabilities for breeding, within a strategic methodology that can be extrapolated to other freshwater wetland ecosystems.

Thus, through tests carried out with different machine learning models, we proceeded to study the levels of association detected among multiple phenotypic variables recorded in each animal raised in freshwater wetlands, along with the substantial trait of “body condition” weighted in them, a well-known characteristic with strong impact on bovine reproductive potential. With powerful computational tools, the 10 most important variables: Phosphatase, Cholesterol, Phosphorus, Hair length, Creatinine, Creatine phosphokinase – CPK, Haematocrit, Body temperature, Haemoglobin and Calcium, all with high correlation to body condition; all measured in situ in the four seasons of the year. From there, using parameters from the A.I. tool (root mean square error – RMSE, mean absolute error – MAE, SHapley Additive exPlanations – SHAP, scaled importance of the variables – SI), a mathematical equation was then developed to incorporate all the weightings obtained per animal, also creating multiplication factors that contribute to obtaining Animal Selection Indicators (ISA), to rank animals with the best adaptive potential to the site, framed in a ranking that serves as a new and innovative protocol in this field. Encircling fundamental concepts of Epigenetics. In this way, the union of bioenvironmental data engineering with powerful computational tools, rural socio-economic concepts, animal welfare and the rational use of environmental resources, can greatly contribute to being more responsible and efficient in key sectors for global food production.

The Journal of Agricultural Science Editorial Highlights are selected by the Editor-in-Chief and are freely available. View the recent selections here.