A computational approach to nutrition science reveals the dynamics of the protein content of human milk

To study the computational aspects of collecting available data in a systematically organized database is becoming a matter of urgency in Nutrition & Food Science. Indeed, major projects on developing big datasets have attempted to fill this gap, but so far with limitations on important facets of food composition such as its temporal variation and uncertainty quantification. The need for methodological data processing, from data acquisition, digital storage, statistics and visualization, via pattern recognition and modelling to prediction and optimization is key to make objective and knowledge-based decisions on scientific and technological issues for food industry, academy and regulation. This study aims to demonstrate the use of a recently developed database on the composition of human milk, the first and easily the most complex food in one's life. We show that the purpose-built ontology of the database, with novelties like considering the food composition as a temporal and stochastic response, can help to recognize patterns in the variation of its protein content. Industrial relevance text This study highlights the need (i) for introducing ISO-like standards how to digitize food composition data; (ii) for computational methods to explore and utilize such databases to their full potentials.