Navigating through Nutritional Landscapes

In contrast to classical hypothesis testing which begins with the formulation of a testable research question, more and more researchers, including nutrition scientists, have been taken a different approach. Given the vast amounts of nutrition and food data being available, data-driven analysis methods are becoming increasingly popular in this field.

In March of this year (2015), a paper was published¹, presenting a comprehensive view of interrelationships between raw foods and nutrients in a network-pattern fashion. The researchers (interestingly with the majority of them having a physics background) generated a food-to-food network of 654 food items which links foods of similar nutritional composition resulting in the formation of hierarchical clusters of food groups (Fig.1). At the highest level two groups can be distinguished, the animal-derived foods group (left) and the plant-derived foods group (right). The underlying reason for this clustering pattern is mainly attributable to the reciprocal contents of protein and carbohydrates in animal and plant products.

Apart from being a simple visualization of already existing nutritional knowledge, a few things stand out. Most importantly, the researchers introduced the term 'nutritional fitness' to identify nutritionally favorable foods. As you may imagine for physicists doing nutrition research, it's a quite technical term. Nutritional fitness measures the frequency by which a single food item is part of an irreducible food set of 4 food items that meets, but do not exceed the daily nutrient demand of an average person. The more often a food item pops up in one of those irreducible food sets after assessing all possible combinations of foods, the higher its nutritional fitness. Almonds, chia seeds, cherimoya (exotic fruit), oecean perch are all foods of high nutritional fitness as indicated by the size of each node in the figure below. In encourage you to read the original article to understand why this is! ;)

Fig 1. The Food-Food Network.

(A–C) Large-scale to small-scale overviews of the network. Each node represents a food, and nodes are connected through links that reflect the similarities between the nutrient contents of foods. In (A–C), each node is colored according to the food category. The size of each node corresponds to the nutritional fitness (NF) of the food. [from publication]

In a very similar manner, the researchers constructed a nutrient-to-nutrient network illustrating positive as well as negative correlations between the abundance of single nutrients across foods (Fig 2.). As was already evident from the food-to-food network protein and carbohydrate contents are inversely correlated. In addition, carbohydrates are also inversely correlated with total lipid content. I am sure you can find correlations that you are particularly interested in! For instance, I am baffled over the positive connection between betaine and trans fats. Betaine is mainly found in beets, other vegetables, cereals and pseudocereals such as quinoa. Trans fats are commonly associated with processed foods, however, this study only involved raw foods. To conclude, networks such as these are very interesting to look at. At first glance, they may just appear to represent common knowledge, but a few moments later, you'll be intrigued by them.

Fig 2. The Nutrient-Nutrient Network.

Each node represents a nutrient, and the nodes are connected through correlations between the abundances of nutrients across all foods. Each node is colored according to the nutrient type. The shape of each node indicates the hierarchical or ‘taxonomic’ level of a nutrient, from ‘Highest’ (a general class of nutrients) to ‘Lowest’ (a specific nutrient). The color and thickness of each link correspond to the sign and magnitude of the correlation, respectively.

1 Citation: Kim S, Sung J, Foo M, Jin Y-S, Kim P-J (2015) Uncovering the Nutritional Landscape of Food. PLoS ONE 10(3): e0118697. doi:10.1371/journal.pone.0118697