Flavor network and the principles of food pairing
Humans have been collecting and cooking food for millions of years, each diet influenced by factors like accessibility, culture, nutritional value, and so on. While there exist approximately 10^6 recipes currently in use, there are around 10^15 total potential recipes. The authors of this paper sought to identify if there exist any quantitative principles behind ingredient pairings, such as two foods that share lots of flavor compounds are more likely to taste well together. They created a flavor network where each node is a different ingredient and the edge between two nodes is proportional in size to the number of flavor compounds shared. The nodes are then categorized into food groups such as fruits, dairy, spices, etc, forming clusters as a result of similar foods sharing more flavor compounds.
By further dividing the network by each ingredient’s prevalence in different regional cuisine, we can see that North American and Western European cuisines tend towards ingredients that share flavor compounds while East Asian and Southern European cuisines avoid recipes with flavor compound sharing ingredients. These differences disappeared when the key ingredients in each cuisine were removed, leading to the “flavor principle” in which a few key ingredients make up the taste palette for a regional cuisine, such as soy sauce in Asian cuisine. These ingredients can create another network demonstrating the similarities between regional cuisines, from which we can observe that Southern European cuisine is much more closely related to Latin American cuisine than Western European cuisine. Do these differences occur because the ingredients available in each cuisine that share lots of flavor compounds tend to pair well together or not? If so, can some sort of positive/negative relationship be established for each region?
As a network, the flavor network is not very dynamic since edges are based on existing flavor compounds in each ingredient and to my knowledge, ingredients don’t tend to gain new flavors out of the blue. Thus, the strong triadic closure property may go unsatisfied if two foods with no shared flavor compounds have strong links to a third food in common. However, if we were to instead consider a new network, one where edges were defined by how well two foods pair together, in a concept known as flavor bridging, these three foods may actually pair well together, satisfying the STCP. Applications of the flavor network may very well explain the strange food combinations that seemingly have no real culinary basis, such as ice cream and fries, or blue cheese and chocolate where the latter food pairing has been shown to share 73 flavor compounds. Since we are only using a fraction of the total number of possible recipes, utilization of the flavor network will hopefully lead to novel food pairings.
Sources: