The ingredient co-occurrence network of packaged foods distributed in the United States

This work presents a novel, comprehensive ingredient co-occurrence network of foods in the United States from the Open Food Database. This network, which contains over 69,000 ingredients and 2 million ingredient co-occurrence relationships, provides a glimpse into the current use of common ingredients for food production and distribution in the United States. Understanding food intake behaviors from this data-driven perspective opens up new avenues for precision health and wellness, for example, in studies of obesity, eating disorders, food allergies, and nursing mothers. The results describe a co-occurrence network with a highly connected core of ingredients found in food distributed in the United States, including salt (4.86% of ingredients), sugar (3.63%), water (3.58%). Further analysis of the most densely connected core of ingredients reveals a majority are added nutrients (niacin, riboflavin) or derived from corn, soy, and wheat. Our results suggest a preference for a small group of ingredients that are used frequently in packaged foods and highlight semantic challenges for aggregating ingredients from food products in a systematic way. This novel approach to analysis of data from freely available food databases can be used to enhance existing methods for modeling food composition at the ingredient level.