Eye on AI - December 13th, 2019
Welcome to Aigora's "Eye on AI" series, where we round up exciting news at the intersection of consumer science and artificial intelligence!
This week, we take a focused look at how AI is improving nutrition, comparing two complementary articles on a Northeastern research group’s attempt to use machine learning to classify food compounds and their relation to health.
How Machine Learning Improves Nutrition by Identifying and Classifying Food Compounds
We begin with an article out of Medical Press, titled “Unraveling the chemical compounds in food could improve how we manage our health,” in which reporter Laura Castañón describes how a group of researchers at Northeastern are using machine learning to better understand food’s complex relationship to human health.
“Looking at databases that are available worldwide, we have between 20 and 30 thousand different chemical compounds in our food,” says Giulia Menichetti, an associate research scientist at Northeastern's Center for Complex Network Research. “And that's just an initial estimation."
Today, the U.S. Department of Agriculture tracks only 150 fats, sugars and other compounds in the things we eat. With so many other biochemical compounds left unaccounted, we’re given a very limited picture of how compounds determine health (perhaps one reason we see so many conflicting diet trends).
The truth is that a food’s nutritional benefit will vary based not only the food itself, but a combination of factors, such as how different compounds found in different foods combine, food preparation, and individual genetics. Albert-László Barabási, network scientist at Northeastern, addresses this issue a related article from Inside Science, titled ‘Studying Food's 'Dark Matter' Could Help Illuminate Diet's Ties to Health.’
"We have all this data coming from genetics, which can explain 10 to 20% of disease causation,” says Barbási. “Where’s the rest? From a person’s environment -- and biggest piece of environment is food…. To address this health crisis, we have to start asking the questions: What exactly is in the food? Which compounds are making us sick, and which are beneficial to us?"
The first step in identifying all this is, of course, identifying and classifying the chemical compounds that make up individual foods. Machine learning makes that possible by sorting through the trove of documents and research papers online, then identifying new compounds and better classifying existing ones. The more the identification and classification stage progresses, the more food recommendations will improve and disease causations understood. Even then, there remain complications.
“... right now, [nutritional information] is calculated using raw, not cooked, food,” notes Nik Sharma, a molecular geneticist turned food writer. “So, if you take an avocado -- a fruit that’s rich in antioxidants -- and apply high temperatures, all the antioxidants get burned off. Similarly, vitamin C is unstable, so cooking a red pepper, for example, will rob it of some of its nutritional value. Even a humble slice of bread has a different nutritional value after it is toasted.”
This research appears to be just in its infancy. But if successful, it could fundamentally alter the ways in which we view and consume food, with the end goal being, as Menichetti notes, “a future in which we can track our daily eating patterns and have a unique description of the chemicals we are ingesting. Combined with our individual genetic variations and health history, this information could provide tangible ways to improve our health, by simply adjusting what we eat.” The question remains whether Northeastern’s researchers will ever be able to combine so many data points effectively, a task that would be all but impossible without advanced knowledge management and machine learning techniques.
That's it for now. If you'd like to receive email updates from Aigora, including weekly video recaps of our blog activity, click on the button below to join our email list. Thanks for stopping by!