Grazing farms differentiation through the expression of microARNs and AI algorithm

Abstract

Milk production based on grazing is being promoted over cattle housed indoors, because of the advantages regarding animal welfare, milk quality and the environment. Cows’ milk is rich in miRNAs, molecules that regulate gene expression in eukaryotes. Their profiles may vary depending on environmental factors such as farm management and feeding. We hypothesize that miRNA can be used as a certification tool for dairy farms whose milk production is based on grazing. The objective is to apply an artificial intelligence algorithm to the results of miRNA expression in milk to evaluate the possibility of designing a fast and cheap traceability tool that can differentiate the milk produced in a grazing-based system from milk produced in indoor systems. Cells and fat fractions were isolated from seventy-three milk tank samples from ‘No-Grazing’ (n=47) vs ‘Grazing’ (n=26) farms. MiRNA expression was analysed in the cells and the fat fractions of the milk samples. Following miRNAs expression analysis, decision trees were built for their expression results using the C4.5 machine learning algorithm. The algorithm was not able to correctly classify each sample in its group, nor was it able to identify relevant miRNAs. We assume that the enormous internal variability (diets, botanical composition of the pastures, and grazing duration, etc.) in commercial grazing farms could be the cause of the difficulty in machine learning of how to classify milk from grazing farms.