How artificial intelligence, AI, can help achieve precision nutrition

Precision nutrition is about better tailoring diets and nutritional recommendations to different people, because one size certainly does not fit all, as I have previously written for Forbes. So to determine the best diet for someone, you just need to figure out what’s going on with that person’s genetics, physiology, microbiome, body type, eating behaviors, stress, social influences, food environment, health status, and all kinds of other things. . that influence nutrition and health. And you have to keep track of how all these things can interact with each other and change over time. No problem, right?

Not really. It can be very complicated to keep track of and figure out all these different things that happen in different ways and at different levels for different people, in different times and circumstances. That’s a lot of ‘differences’. But nowadays, when you have to figure out something very complicated, you have a potential friend in AI, that is, artificial intelligence.

A major challenge is that science hasn’t even figured out how all these different factors interact and influence the way a person’s diet affects his or her health. Certainly, studies to date have provided insights into how each of these factors may work individually and for certain types of people. But combining these insights is another matter and many gaps remain.

That’s because a single traditional real-world laboratory, clinical or epidemiological study alone cannot explain, measure and track all the different factors and outcomes for all kinds of people. No matter how hard you try to design the “perfect” study, you will undoubtedly fail to include all types of people and measure all relevant factors and outcomes.

Furthermore, even if you designed the “perfect” study, you would have to wait a very long time before you got all the results you needed. It can take years, even decades, for the effects of diet to manifest in various health problems. Anyone who ate like garbage and considered ketchup a vegetable until they were 20 will tell you that.

So if you really want to know how to do precision nutrition, you have to somehow combine data from many different studies and fill in the gaps. You also want to find ways to extend the results of a particular study to people who did not participate in that study and to circumstances that were not covered. All of this can be far too complex for any one person or even a team of people to do without help.

Enter AI and play Randy Newman’s song, “You’ve Got A Friend in Me.” Such computer-aided techniques can keep track of many different things, combining different data sets in different ways and figuring out how they fit together. These techniques can also determine how the results of a single nutritional study may apply to different conditions and situations, thereby increasing the usefulness and value of that study. And several AI techniques can do this quickly, much faster than humans. These are just some of the ways AI can help achieve precision nutrition.

To understand how AI can do these things, you first need to know what AI is. Today, AI has become such a sexy term that people can use it without even knowing what the term means, saying things like, “Hey, can you do this AI?” AI is an umbrella term that basically encompasses any computer-aided technique that can replicate something a human brain would normally do, beyond simply following step-by-step directions. An AI approach can therefore assess situations or make decisions independently. There are already many different types of AI approaches, methods and tools and the list continues to grow every year.

One way to classify AI techniques is based on a continuum of how these techniques are designed and work. On one side are purely data-driven AI approaches. These are ‘top-down’ techniques that start with a collection of data and try to identify patterns, trends and associations from this data. It’s a bit like the way a statistician might analyze a series of data. But the AI ​​algorithm can do it much faster and can perform many different analyzes on multiple data sets at the same time.

Let’s look at a theoretical example. A data-driven AI approach can analyze different data sets, break down the data in different ways, and discover that people who eat a particular food item tend to live longer. Let’s call this food item a “Swift Bieber,” a completely fictional term named after nothing in particular. The AI ​​algorithm may be able to associate Swift Biebers with a longer lifespan, but cannot explain why this association actually exists. It cannot really distinguish whether Swift Bieber’s consumption has an actual beneficial nutritional effect or whether it is a coincidence. Perhaps those who tend to eat Swift Biebers at the same time also tend to eat some other food item not included in the data set that actually provides the right solution. Or maybe people who are less stressed are more likely to have the time and money to eat Swift Biebers. Swift Biebers could actually be a red herring, meaning something misleading or distracting instead of something made out of fish.

At the other end of the spectrum are mechanistic or explainable AI approaches. These AI methods attempt to recreate what actually happens from the bottom up, by recreating the actual mechanisms behind a process or decision. They are considered explainable because you know the specific reasons why a result was generated.

This is analogous to what scientists do when they design experiments in a laboratory to test what might happen. The difference is that the AI ​​algorithm or model is not limited to a physical laboratory and can serve as a “virtual laboratory” representing an entire person, a group of people, a population, or an entire geographic area. The model can then perform experiments in the ‘safety’ of a computer in ways that in real life would be too complex, too expensive, too time-consuming, too impractical or even too dangerous. The mechanistic AI tool could then use the results of these experiments to make subsequent recommendations, much like a human runs thought experiments in his or her head before taking action.

For example, a mechanistic AI approach could involve displaying the different reasons why someone chooses to eat a Swift Bieber. It could also show the different nutrients in a Swift Bieber, how they are broken down in the body, how these nutrients then affect different organs and how this ultimately affects lifespan. Then, this AI model could look at what would happen over time if different people ate Swift Biebers and decide who would benefit from eating Swift Biebers and how.

These different AI techniques across the spectrum can work together and also be integrated. A purely data-driven approach can suggest associations (for example, look at Swift Biebers) that can guide the construction of more mechanistic AI approaches (for example, let’s figure out what Swift Biebers actually do with the body). Likewise, a mechanistic AI approach can help determine where data-driven approaches are needed. Let’s say you’re trying to pinpoint the mechanisms by which Swift Bieber influences the microbiome, but you can’t pinpoint them because there are no traditional studies that clearly reveal relationships, patterns, and trends. Therefore, it may be useful for data-driven AI approaches to sift through this microbiome data.

Of course, you shouldn’t automatically trust everything AI tells you. Just as a poorly designed clinical trial or observational study can lead to misleading results, so can a poorly designed AI approach. That’s why you need to know what’s under the hood of an AI approach and understand its relative strengths and weaknesses. At the same time, no AI approach – like no real-world research – will be perfect. Don’t let the perfect be the enemy of the good and let the imperfections of an AI approach keep you from using it out of risk aversion.

Incorporating more AI and other computer-aided approaches to make more precise recommendations is not entirely new and has already been done in other areas. Fields such as meteorology, finance and aerospace engineering have long used computer-aided techniques to bring together and analyze complex data from disparate sources to generate more accurate insights and forecasts.

So while AI is unlikely to conflict with some of the already established nutritional insights, such as the value of eating fruits and vegetables, the nutrition field is ripe for change. There are too many people claiming that this and that super-duper-just-eat-this diet works for everyone. But not everyone is the same and has the same circumstances, and that is exactly the problem. Achieving more precision nutrition is not easy. but you have a potential friend in AI. But like any potential boyfriend, you need to treat him well and know what he can and cannot do.