AI and you: Is the promise of personalized nutrition apps worth the hype?
As a type 2 diabetic, Michael Snyder has long been interested in how blood sugar levels vary from one person to another in response to the same food, and whether a more personalized approach to nutrition could help tackle the rapidly cascading levels of diabetes and obesity in much of the western world.
Eight years ago, Snyder, who directs the Center for Genomics and Personalized Medicine at Stanford University, decided to put his theories to the test. In the 2000s continuous glucose monitoring, or CGM, had begun to revolutionize the lives of diabetics, both type 1 and type 2. Using spherical sensors which sit on the upper arm or abdomen – with tiny wires that pierce the skin – the technology allowed patients to gain real-time updates on their blood sugar levels, transmitted directly to their phone.
It gave Snyder an idea for his research at Stanford. Applying the same technology to a group of apparently healthy people, and looking for ‘spikes’ or sudden surges in blood sugar known as hyperglycemia, could provide a means of observing how their bodies reacted to an array of foods.
“We discovered that different foods spike people differently,” he says. “Some people spike to pasta, others to bread, others to bananas, and so on. It’s very personalized and our feeling was that building programs around these devices could be extremely powerful for better managing people’s glucose.”
Unbeknown to Snyder at the time, thousands of miles away, a group of Israeli scientists at the Weizmann Institute of Science were doing exactly the same experiments. In 2015, they published a landmark paper which used CGM to track the blood sugar levels of 800 people over several days, showing that the biological response to identical foods can vary wildly. Like Snyder, they theorized that giving people a greater understanding of their own glucose responses, so they spend more time in the normal range, may reduce the prevalence of type 2 diabetes.
The commercial potential of such apps is clear, but the underlying science continues to generate intriguing findings.
“At the moment 33 percent of the U.S. population is pre-diabetic, and 70 percent of those pre-diabetics will become diabetic,” says Snyder. “Those numbers are going up, so it’s pretty clear we need to do something about it.”
Fast forward to 2022,and both teams have converted their ideas into subscription-based dietary apps which use artificial intelligence to offer data-informed nutritional and lifestyle recommendations. Snyder’s spinoff, January AI, combines CGM information with heart rate, sleep, and activity data to advise on foods to avoid and the best times to exercise. DayTwo–a start-up which utilizes the findings of Weizmann Institute of Science–obtains microbiome information by sequencing stool samples, and combines this with blood glucose data to rate ‘good’ and ‘bad’ foods for a particular person.
“CGMs can be used to devise personalized diets,” says Eran Elinav, an immunology professor and microbiota researcher at the Weizmann Institute of Science in addition to serving as a scientific consultant for DayTwo. “However, this process can be cumbersome. Therefore, in our lab we created an algorithm, based on data acquired from a big cohort of people, which can accurately predict post-meal glucose responses on a personal basis.”
The commercial potential of such apps is clear. DayTwo, who market their product to corporate employers and health insurers rather than individual consumers, recently raised $37 million in funding. But the underlying science continues to generate intriguing findings.
Last year, Elinav and colleagues published a study on 225 individuals with pre-diabetes which found that they achieved better blood sugar control when they followed a personalized diet based on DayTwo’s recommendations, compared to a Mediterranean diet. The journal Cell just released a new paper from Snyder’s group which shows that different types of fibre benefit people in different ways.
“The idea is you hear different fibres are good for you,” says Snyder. “But if you look at fibres they’re all over the map—it’s like saying all animals are the same. The responses are very individual. For a lot of people [a type of fibre called] arabinoxylan clearly reduced cholesterol while the fibre inulin had no effect. But in some people, it was the complete opposite.”
Eight years ago, Stanford's Michael Snyder began studying how continuous glucose monitors could be used by patients to gain real-time updates on their blood sugar levels, transmitted directly to their phone.
The Snyder Lab, Stanford Medicine
Because of studies like these, interest in precision nutrition approaches has exploded in recent years. In January, the National Institutes of Health announced that they are spending $170 million on a five year, multi-center initiative which aims to develop algorithms based on a whole range of data sources from blood sugar to sleep, exercise, stress, microbiome and even genomic information which can help predict which diets are most suitable for a particular individual.
“There's so many different factors which influence what you put into your mouth but also what happens to different types of nutrients and how that ultimately affects your health, which means you can’t have a one-size-fits-all set of nutritional guidelines for everyone,” says Bruce Y. Lee, professor of health policy and management at the City University of New York Graduate School of Public Health.
With the falling costs of genomic sequencing, other precision nutrition clinical trials are choosing to look at whether our genomes alone can yield key information about what our diets should look like, an emerging field of research known as nutrigenomics.
The ASPIRE-DNA clinical trial at Imperial College London is aiming to see whether particular genetic variants can be used to classify individuals into two groups, those who are more glucose sensitive to fat and those who are more sensitive to carbohydrates. By following a tailored diet based on these sensitivities, the trial aims to see whether it can prevent people with pre-diabetes from developing the disease.
But while much hope is riding on these trials, even precision nutrition advocates caution that the field remains in the very earliest of stages. Lars-Oliver Klotz, professor of nutrigenomics at Friedrich-Schiller-University in Jena, Germany, says that while the overall goal is to identify means of avoiding nutrition-related diseases, genomic data alone is unlikely to be sufficient to prevent obesity and type 2 diabetes.
“Genome data is rather simple to acquire these days as sequencing techniques have dramatically advanced in recent years,” he says. “However, the predictive value of just genome sequencing is too low in the case of obesity and prediabetes.”
Others say that while genomic data can yield useful information in terms of how different people metabolize different types of fat and specific nutrients such as B vitamins, there is a need for more research before it can be utilized in an algorithm for making dietary recommendations.
“I think it’s a little early,” says Eileen Gibney, a professor at University College Dublin. “We’ve identified a limited number of gene-nutrient interactions so far, but we need more randomized control trials of people with different genetic profiles on the same diet, to see whether they respond differently, and if that can be explained by their genetic differences.”
Some start-ups have already come unstuck for promising too much, or pushing recommendations which are not based on scientifically rigorous trials. The world of precision nutrition apps was dubbed a ‘Wild West’ by some commentators after the founders of uBiome – a start-up which offered nutritional recommendations based on information obtained from sequencing stool samples –were charged with fraud last year. The weight-loss app Noom, which was valued at $3.7 billion in May 2021, has been criticized on Twitter by a number of users who claimed that its recommendations have led to them developed eating disorders.
With precision nutrition apps marketing their technology at healthy individuals, question marks have also been raised about the value which can be gained through non-diabetics monitoring their blood sugar through CGM. While some small studies have found that wearing a CGM can make overweight or obese individuals more motivated to exercise, there is still a lack of conclusive evidence showing that this translates to improved health.
However, independent researchers remain intrigued by the technology, and say that the wealth of data generated through such apps could be used to help further stratify the different types of people who become at risk of developing type 2 diabetes.
“CGM not only enables a longer sampling time for capturing glucose levels, but will also capture lifestyle factors,” says Robert Wagner, a diabetes researcher at University Hospital Düsseldorf. “It is probable that it can be used to identify many clusters of prediabetic metabolism and predict the risk of diabetes and its complications, but maybe also specific cardiometabolic risk constellations. However, we still don’t know which forms of diabetes can be prevented by such approaches and how feasible and long-lasting such self-feedback dietary modifications are.”
Snyder himself has now been wearing a CGM for eight years, and he credits the insights it provides with helping him to manage his own diabetes. “My CGM still gives me novel insights into what foods and behaviors affect my glucose levels,” he says.
He is now looking to run clinical trials with his group at Stanford to see whether following a precision nutrition approach based on CGM and microbiome data, combined with other health information, can be used to reverse signs of pre-diabetes. If it proves successful, January AI may look to incorporate microbiome data in future.
“Ultimately, what I want to do is be able take people’s poop samples, maybe a blood draw, and say, ‘Alright, based on these parameters, this is what I think is going to spike you,’ and then have a CGM to test that out,” he says. “Getting very predictive about this, so right from the get go, you can have people better manage their health and then use the glucose monitor to help follow that.”
Podcast: The Friday Five weekly roundup in health research
The Friday Five covers five stories in health research that you may have missed this week. There are plenty of controversies and troubling ethical issues in science – and we get into many of them in our online magazine – but this news roundup focuses on scientific creativity and progress to give you a therapeutic dose of inspiration headed into the weekend.
Covered in this week's Friday Five:
- Sex differences in cancer
- Promising research on a vaccine for Lyme disease
- Using a super material for brain-like devices
- Measuring your immunity to Covid
- Reducing dementia risk with leisure activities
One day in recent past, scientists at Columbia University’s Creative Machines Lab set up a robotic arm inside a circle of five streaming video cameras and let the robot watch itself move, turn and twist. For about three hours the robot did exactly that—it looked at itself this way and that, like toddlers exploring themselves in a room full of mirrors. By the time the robot stopped, its internal neural network finished learning the relationship between the robot’s motor actions and the volume it occupied in its environment. In other words, the robot built a spatial self-awareness, just like humans do. “We trained its deep neural network to understand how it moved in space,” says Boyuan Chen, one of the scientists who worked on it.
For decades robots have been doing helpful tasks that are too hard, too dangerous, or physically impossible for humans to carry out themselves. Robots are ultimately superior to humans in complex calculations, following rules to a tee and repeating the same steps perfectly. But even the biggest successes for human-robot collaborations—those in manufacturing and automotive industries—still require separating the two for safety reasons. Hardwired for a limited set of tasks, industrial robots don't have the intelligence to know where their robo-parts are in space, how fast they’re moving and when they can endanger a human.
Over the past decade or so, humans have begun to expect more from robots. Engineers have been building smarter versions that can avoid obstacles, follow voice commands, respond to human speech and make simple decisions. Some of them proved invaluable in many natural and man-made disasters like earthquakes, forest fires, nuclear accidents and chemical spills. These disaster recovery robots helped clean up dangerous chemicals, looked for survivors in crumbled buildings, and ventured into radioactive areas to assess damage.
Now roboticists are going a step further, training their creations to do even better: understand their own image in space and interact with humans like humans do. Today, there are already robot-teachers like KeeKo, robot-pets like Moffin, robot-babysitters like iPal, and robotic companions for the elderly like Pepper.
But even these reasonably intelligent creations still have huge limitations, some scientists think. “There are niche applications for the current generations of robots,” says professor Anthony Zador at Cold Spring Harbor Laboratory—but they are not “generalists” who can do varied tasks all on their own, as they mostly lack the abilities to improvise, make decisions based on a multitude of facts or emotions, and adjust to rapidly changing circumstances. “We don’t have general purpose robots that can interact with the world. We’re ages away from that.”
Robotic spatial self-awareness – the achievement by the team at Columbia – is an important step toward creating more intelligent machines. Hod Lipson, professor of mechanical engineering who runs the Columbia lab, says that future robots will need this ability to assist humans better. Knowing how you look and where in space your parts are, decreases the need for human oversight. It also helps the robot to detect and compensate for damage and keep up with its own wear-and-tear. And it allows robots to realize when something is wrong with them or their parts. “We want our robots to learn and continue to grow their minds and bodies on their own,” Chen says. That’s what Zador wants too—and on a much grander level. “I want a robot who can drive my car, take my dog for a walk and have a conversation with me.”
Columbia scientists have trained a robot to become aware of its own "body," so it can map the right path to touch a ball without running into an obstacle, in this case a square.
Jane Nisselson and Yinuo Qin/ Columbia Engineering
Today’s technological advances are making some of these leaps of progress possible. One of them is the so-called Deep Learning—a method that trains artificial intelligence systems to learn and use information similar to how humans do it. Described as a machine learning method based on neural network architectures with multiple layers of processing units, Deep Learning has been used to successfully teach machines to recognize images, understand speech and even write text.
Trained by Google, one of these language machine learning geniuses, BERT, can finish sentences. Another one called GPT3, designed by San Francisco-based company OpenAI, can write little stories. Yet, both of them still make funny mistakes in their linguistic exercises that even a child wouldn’t. According to a paper published by Stanford’s Center for Research on Foundational Models, BERT seems to not understand the word “not.” When asked to fill in the word after “A robin is a __” it correctly answers “bird.” But try inserting the word “not” into that sentence (“A robin is not a __”) and BERT still completes it the same way. Similarly, in one of its stories, GPT3 wrote that if you mix a spoonful of grape juice into your cranberry juice and drink the concoction, you die. It seems that robots, and artificial intelligence systems in general, are still missing some rudimentary facts of life that humans and animals grasp naturally and effortlessly.
How does one give robots a genome? Zador has an idea. We can’t really equip machines with real biological nucleotide-based genes, but we can mimic the neuronal blueprint those genes create.
It's not exactly the robots’ fault. Compared to humans, and all other organisms that have been around for thousands or millions of years, robots are very new. They are missing out on eons of evolutionary data-building. Animals and humans are born with the ability to do certain things because they are pre-wired in them. Flies know how to fly, fish knows how to swim, cats know how to meow, and babies know how to cry. Yet, flies don’t really learn to fly, fish doesn’t learn to swim, cats don’t learn to meow, and babies don’t learn to cry—they are born able to execute such behaviors because they’re preprogrammed to do so. All that happens thanks to the millions of years of evolutions wired into their respective genomes, which give rise to the brain’s neural networks responsible for these behaviors. Robots are the newbies, missing out on that trove of information, Zador argues.
A neuroscience professor who studies how brain circuitry generates various behaviors, Zador has a different approach to developing the robotic mind. Until their creators figure out a way to imbue the bots with that information, robots will remain quite limited in their abilities. Each model will only be able to do certain things it was programmed to do, but it will never go above and beyond its original code. So Zador argues that we have to start giving robots a genome.
How does one do that? Zador has an idea. We can’t really equip machines with real biological nucleotide-based genes, but we can mimic the neuronal blueprint those genes create. Genomes lay out rules for brain development. Specifically, the genome encodes blueprints for wiring up our nervous system—the details of which neurons are connected, the strength of those connections and other specs that will later hold the information learned throughout life. “Our genomes serve as blueprints for building our nervous system and these blueprints give rise to a human brain, which contains about 100 billion neurons,” Zador says.
If you think what a genome is, he explains, it is essentially a very compact and compressed form of information storage. Conceptually, genomes are similar to CliffsNotes and other study guides. When students read these short summaries, they know about what happened in a book, without actually reading that book. And that’s how we should be designing the next generation of robots if we ever want them to act like humans, Zador says. “We should give them a set of behavioral CliffsNotes, which they can then unwrap into brain-like structures.” Robots that have such brain-like structures will acquire a set of basic rules to generate basic behaviors and use them to learn more complex ones.
Currently Zador is in the process of developing algorithms that function like simple rules that generate such behaviors. “My algorithms would write these CliffsNotes, outlining how to solve a particular problem,” he explains. “And then, the neural networks will use these CliffsNotes to figure out which ones are useful and use them in their behaviors.” That’s how all living beings operate. They use the pre-programmed info from their genetics to adapt to their changing environments and learn what’s necessary to survive and thrive in these settings.
For example, a robot’s neural network could draw from CliffsNotes with “genetic” instructions for how to be aware of its own body or learn to adjust its movements. And other, different sets of CliffsNotes may imbue it with the basics of physical safety or the fundamentals of speech.
At the moment, Zador is working on algorithms that are trying to mimic neuronal blueprints for very simple organisms—such as earthworms, which have only 302 neurons and about 7000 synapses compared to the millions we have. That’s how evolution worked, too—expanding the brains from simple creatures to more complex to the Homo Sapiens. But if it took millions of years to arrive at modern humans, how long would it take scientists to forge a robot with human intelligence? That’s a billion-dollar question. Yet, Zador is optimistic. “My hypotheses is that if you can build simple organisms that can interact with the world, then the higher level functions will not be nearly as challenging as they currently are.”
Lina Zeldovich has written about science, medicine and technology for Popular Science, Smithsonian, National Geographic, Scientific American, Reader’s Digest, the New York Times and other major national and international publications. A Columbia J-School alumna, she has won several awards for her stories, including the ASJA Crisis Coverage Award for Covid reporting, and has been a contributing editor at Nautilus Magazine. In 2021, Zeldovich released her first book, The Other Dark Matter, published by the University of Chicago Press, about the science and business of turning waste into wealth and health. You can find her on http://linazeldovich.com/ and @linazeldovich.