Machine Intelligence Applied to Sustainability: A Systematic Methodological Proposal to Identify Sustainable Animals

We consider sustainable animals those which are well-adapted to the environment, productive, feed efficient in converting foods into animal products (milk, meat, wool and honey), clinically healthy and presenting a low carbon and water footprints, having their animal welfare assured. Methodologies to identify these animals are necessary, though non-existent. Here, we develop a systematic methodological approach using several machine learning techniques to identify sustainable animals through phenotypic biomarkers for sustainable livestock production systems. Thermoregulatory, morphological, haematological, serum biochemical, hormonal and productive responses, carcass traits and feed efficiency measures of 62 bulls were collected and analyzed for this study. The database contains 40 variables with 10,354 pieces of information. We identified, through machine learning, a typology for sustainable animals, which were bulls with sustainable resilience and adaptive capacity to the environment, productive sustainability with heavy carcasses and high yield and feed sustainability with negative RFI. The main phenotypic biomarkers to identify sustainable animals are rectal temperature, mean corpuscular volume and white blood cells – sustainable resilience and adaptive capacity; precocity, birth weight and carcass yield – productive sustainability; and residual feed intake – feed sustainability. We recommend using this methodological proposal to identify sustainable animals, in which they consume less food, are adapted to the environment and are more productive, increasing the farm's income and possibly reducing the water and carbon footprints.