Artificial Intelligence and Dietary Supplements: The Future of Formulating?

Jun 7, 2018
  • Artificial intelligence (AI) seems galaxies away from the world of dietary supplement manufacturing. The idea that computer systems, by simulating human intelligence, can play a role in creating dietary supplement made from herbs, vitamins, and other ingredients seems far-fetched. Yet, according to one prominent supplement company, Life Extension, AI will be the future of dietary supplement creation, helping companies not only formulate products and bring them to market more quickly, but also helping them determine which ingredients to study and how to study them.

    Nutritional Outlook spoke with Michael Smith, MD, Life Extension’s senior scientist, about how Life Extension is already using AI in its dietary supplements business. The company’s mission is to develop longevity products to help consumers live longer, healthier lives. In the following slides, Smith discusses how supplement makers can harness the power of AI.



    Photo ©

  • Picking Ingredients

    Smith says artificial intelligence is helping Life Extension researchers more efficiently pinpoint which ingredients to focus their attention on in order to formulate a more efficacious product.

    For instance, he says, Life Extension recently launched a dietary supplement to help the body support itself against the effects of cellular senescence. (Cellular senescence refers to what happens when cells stop dividing and eventually die as a natural part of aging.) Using AI algorithms, Life Extension researchers were able to generate a list of 70 potentially effective nutrients to combat cellular senescence, Smith says.

    He continues: “We then asked the computer to rank those nutrients based on the type of research it was looking at, and we then took from that ranked list the top four nutrients. Those basically became our first [AI-formulated supplement] product. So, we were able to formulate a cellular senescence product with the top four nutrients that affect senescent cells the best.”

    Photo ©

  • Feeding the Machine

    How does an AI engine make human-like judgments? To start with, by using background information it's already been “fed."

    For instance, Smith says, Life Extension might upload to the AI engine information the company has garnered from scientific literature and from its own research about some of the different pathways in human cells that are pro-aging and antiaging.

    “There are thousands of these pathways, and they all interact in very complex ways,” he explains. “To understand how these pathways work, to really understand why a cell ages—I mean, again, you’re talking about years and years of research. We would have to have hundreds of teams of researchers looking at that every day for decades just to understand these pathways and their interaction.”

    Instead, he says, this is where AI comes in. “We’re able to feed into the AI engine a few pathways that we do feel confident about in terms of knowing how they work, and then based on those inputs, AI technology can start ‘understanding’ all these other pathways. It can actually learn and then give us an output that shows, in detail, how all these pathways interact,” he says. “And then it can go further. It can scour all the research literature and say, ‘Based on all these pathways I’m looking at, and based on all the research on nutraceuticals like pomegranate, resveratrol, green tea, etc., I think these are the four nutrients you should look at because these four nutrients seem to affect these aging pathways in a way that actually will lead to healthy longevity in the future.”

    It’s not necessary to feed the artificial intelligence engine every single existing study or piece of data either, Smith adds; rather, the AI engine will generate its output based on algorithms.

    He compares the process to the process used by facial-recognition software. “You don’t need to feed a facial-recognition system every single face possible. That defeats the purpose, right? And that would be hard to do. Instead, you simply teach it some commonalities about human faces, like about different nose sizes. Maybe you show it a small nose and a large nose. You show white skin and brown skin. You show dark hair and light hair. These are basic, common properties of a human face. You feed that in, and then the AI technology is able to rearrange that because it’s learning these commonalities, and it starts putting faces together—and, in the end, it does result in a human face.”



    Photo © Film

  • AI Is Not a Search Engine 

    Although artificial intelligence engines can be told to sift through and generate results from copious amounts of scientific literature, AI is much more than just a search engine, Smith says.

    “I can Google, ‘Does green tea affect cellular senescence?’ and I probably would get the results of a few studies back,” he says. “But Google is simply bringing back things that it found by looking for keywords. It’s not looking at these results from an analytical point of view. It’s not really utilizing any sort of metrics to pick the best-quality results. It’s just giving me a keyword-based results page.”

    By contrast, Smith says, AI not only sifts through the data to find applicable studies, it can predict which studies are likely to produce the end results scientists are looking for. In short, he says, “It’s making better predictions about which nutrients most likely will have those long-term effects, without actually requiring us to study them for decades,” he says.

    “It is actually analyzing the data in the same way a human brain does,” he continues. “It’s looking at the statistical models. It’s looking at how the results were done. So, when it brings back a results page, it has actually analyzed those studies just like a human brain does and said, ‘Based on what I’ve already learned using the input that you gave me in the beginning, I was able to learn what those pathways were doing and I can tell you that these three studies are the three you really want to look at because they’re showing the right pathways in the right way.”

    “The computer is actually analyzing that initial search, if that makes sense,” he says.


    Photo ©

  • Saving Time

    By cutting through the time it takes to analyze a vast amount of data, artificial intelligence can help researchers develop products much more quickly, Smith says.

    Life Extension, for instance, used AI to create the aforementioned cellular senescence supplement, whittling down the list of 70 potential nutrients to four final ingredients in the product. “If I were doing that research myself in a lab, [the product’s creation] would be decades away,” he adds. “We sped up time using AI technology by at least a couple of decades.”

    Life Extension worked with medical technology firm Insilico Medicine Inc. (Rockville, MD) and Insilico’s artificial intelligence technology, which the company uses for drug, biomarker, and aging research.

    AI is also helping Life Extension better determine which ingredients to study moving forward. Without the guidance of AI, Smith says, “It would take decades to actually study whether a group of nutrients actually had longevity effects at the cell level—and then ultimately at the organism level. You’d have to follow people for a very long time.”

    He continues, “Let’s say I wanted to test whether resveratrol has an antiaging effect on a certain population of patients with diabetes. Well, that would require a huge study, millions of dollars, and many years. And we might not even know if we’d get a positive result. We’d have a hypothesis, but we would have to invest a lot really not knowing what the outcome would be. But what AI allows me to do is, I can put a lot of this information into an AI engine that then can almost run simulations of these trials, and I can get an idea of whether it would be worth it to invest money in an actual trial” because AI can indicate which research might yield positive results.

    For instance, he says, “Maybe the engine comes back and says, ‘Well, resveratrol in that population might not have that effect; however, if you studied it in people with heart failure, it would tell us the pathway to go down where we could be the most effective. That’s really exciting.”

    The ability to better determine where to invest research dollars is especially important to the nutraceutical industry. While drug makers have millions of dollars to invest in clinical research, nutraceutical researchers have much smaller budgets; therefore, being able to better utilize those research dollars is extremely beneficial, Smith says. AI, he says, is “actually helping us be very effective with our research money.”


    Photo ©


  • Artificial Intelligence Is Just the Starting Point

    Ideally, companies will continue pursuing the gold standard—human clinical trials—on any ingredients and products initially developed using artificial intelligence. For instance, Smith says, once Life Extension created its cellular senescence supplement using the four ingredients “picked” by AI—epigallocatechin gallate (EGCG) from green tea; myricetin, an antioxidant from citrus fruit; NAC (N-acetyl-L-cysteine); and alpha-tocotrienols—the company followed up by performing in vitro studies of its own on the select ingredients to verify the outcomes predicted by the AI engine.

    “Before we formulated the product, the first thing we did is verify, in a lab, those pathways that the AI engine told us this formula would work on. And that’s the beauty of it,” he says. Without AI, he notes, it would have taken Life Extension many years to determine which pathways each of those four ingredients might impact. In addition, he says, the AI engine predicted the ideal dosage combination of all four ingredients.

    Once Life Extension’s own in vitro data seemed to back the AI conclusions, the company proceeded with formulating the product. Following in vitro studies, Smith says, the next step for Life Extension will be running human clinical trials. “We’ll hopefully incorporate some of that other AI information to tell us which population we should test this in and how long we should test it to really get the results we want,” he adds.

    “AI is allowing us to get to the lab faster to verify, and eventually will allow us to get to clinical research quicker as well,” he says.


    Photo ©

  • Selling an AI Product

    Life Extension has embraced a marriage between artificial intelligence and dietary supplements, but have its customers done the same? Smith says yes.

    Life Extension formulated its GeroProtect supplement line around AI technology. The first product to debut in the line was the cellular senescence supplement discussed earlier, called GeroProtect Ageless Cell, which the company markets as “defying aging with nutrients identified by artificial intelligence.”

    On its website, Life Extension states: “The Ageless Cell formula was created through a unique partnership with Insilico Medicine Inc., a medical technology company who created a database of 200+ geroprotectors shown to extend life span, then used a proprietary algorithm based on 21st century artificial intelligence computing techniques to score these compounds’ impact on pathways associated with normal aging. Without Insilico's A.I., this process could have taken [ital] decades.”

    Smith says consumer response to the product and AI have been “quite positive.”
    “Does AI have a weird connotation to people, or a negative one?” he asks. “It turns out, no. People understand that AI technology is already out there.” Consumers, he says, “are actually interested in what it really means and how it really works. So, it’s been very successful for us.” In fact, he says, the second product in the GeroProtect line features the word AI right in its name: GeroProtect Longevity A.I.


    Photo from Life Extension

  • The Future

    Smith says we’re still at the forefront of using artificial intelligence to benefit dietary supplement formulators. He also admits that even as Life Extension pushes forward with developing products using AI, the company will also continue formulating many products more traditionally and without the use of AI—e.g., drawing from the company’s existing ingredient knowledge and existing science to pick ingredients and formulate products.

    Smith has a very positive view on how AI stands to change the future of the supplements industry. “It’s going to change everything because we’re going to be able to bring to market products that have evidence-based research behind them, to help with all kinds of degenerative diseases of aging,” he says.

    “We’re going to be able to bring those products to market at an extremely fast rate, much faster than we were able to before,” he says. “As I said, there’s a time gap between what we start to understand in the lab versus what we can study clinically. We’re bridging that gap by decades now by using AI technology, so you could potentially see an explosion of effective and safe products against many age-related disorders in just the next few years, whereas without AI technology, we’re talking decades.”


    Photo ©