Vitafoods Insights is part of the Informa Markets Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Fast-tracking ‘transformative’ ingredient discovery: How AI is changing new product development

Article-Fast-tracking ‘transformative’ ingredient discovery: How AI is changing new product development

© AdobeStock/elenabsl Fast-tracking ‘transformative’ ingredient discovery: How AI is changing new product development
While artificial intelligence (AI) has become something of an overused buzzword in the nutraceutical world, true AI has the power to allow R&D scientists to discover genuinely transformative ingredients that would take millions of years to develop with traditional methods. “There is no other way forward in this space,” says Nora Khaldi, CEO of Nuritas.

In recent years, investors have seen growing numbers of companies interested in using AI and generative AI to discover new food ingredients when previously, they would have entered the pharmaceutical space, said Gil Horsky, founding partner at venture capital fund Flora Ventures, at a Future Food Tech panel discussion last month.

So, is the food and nutraceutical industry about to experience a wave of companies developing new ingredients using AI?

According to Ehsan Moaseri, CEO and founder at Nulixir, a US startup developing nanocarriers for functional ingredients, many people are using the term AI very loosely without really understanding what it is. Others incorrectly label a process as AI when, in fact, it is based on neural network machines.

“If it is actual AI, it’s definitely a fascinating platform to have,” Moaseri said. “It simply enables you to gain significantly more knowledge about any topic, [such as] fortification and bioavailability in a significantly shorter amount of time. But applying the principles and assumptions behind these AI models [is] sometimes very challenging. No matter how strong your hardware is, your output data is only as good as the input data you provided.

“One of the key challenges […] we see with AI models is that in many of those cases, you start creating very major claims from your output data. Well, you haven’t really screened that input data to make sure you can actually back those claims that you are now creating...”

Verifying the robustness of claims that are being made is an area that requires greater regulation, Moaseri said, adding: “I see that as a major challenge compared to what we have in pharma.”

Understanding molecules and associating them with a health benefit

One of the first companies to use AI in ingredient development for food agriculture applications was Irish company Nuritas. Its CEO and founder, Khaldi, was clear about the role this technology has to play in ingredient R&D. “There is no other way forward in this space – creating new ingredients, whether it’s in pharma or whether it’s in food – without today the use of AI,” she said.

According to Khaldi, AI has the most potential within food product development when applied at a molecular – rather than diet – level.

Understanding the molecules within a source material and associating those molecules to a health benefit – that’s what AI can do. And that’s the exact same problem pharma is solving,” she said.

“How do we associate a molecule – no matter what that molecule is, whether it’s in a natural product or whether it’s synthesised by human – […] with a human and what can it do for the human? Is it beneficial? Is it not? Is it neutral?”

AI can speed up NPD – but years of data collection are still necessary

Khaldi noted, however, that while AI can help solve these fundamental questions, it does require years of experiments to get there. Nuritas, for instance, spent seven years experimentally testing short protein peptides and different cell types from human body tissues and studying the data from a molecular perspective.

“We built in the lab digestive systems where we followed the molecule through the digestive system and we did thousands, millions of those experiments to just build a predictor in the bioavailability space,” she said. “Forget efficacy – efficacy is a whole different world.

“Then, when the AI works, it starts from: what do we actually need? We need an ingredient that’s going to be [generally recognised as safe] GRAS, natural, but also efficacious, and also orally available, and also heat stable. All those are factors [related to] the molecule itself and can only be done through AI.”

Companies must be prepared to invest the time and resources required to collect and develop robust data.

[You cannot] pluck it out of the internet,” she added. “You have to go into the lab and produce it, and it’s costly and it takes time. But once you get there, you can create ingredients that are transformative.”

Can AI-powered products be affordable?

Carole Bingley from RSSL voiced concerns over the affordability of AI-powered food and drink ingredients. Creating ingredients that have undergone extensive, pharmaceutical-level testing could create cost barriers, pricing the healthiest products out of reach of the people who may stand to benefit from them the most, she said.

According to Khaldi, however, by speeding up the bioactive or ingredient discovery phase, AI can ultimately reduce the cost of R&D and product development. Nuritas recently calculated that it would have taken 30 million years to identify and develop one of its ingredients, PeptiStrong, with the traditional methods of discovery used today.

When developing health ingredients, food manufacturers must factor in not only efficacy of the molecule but also taste and texture, which can complicate the process and make product development very expensive. AI can take these parameters into account from the very beginning.

That’s not to say, however, that AI helps product developers create the perfect product from the very start. Khaldi noted that Nuritas’ first iteration of this ingredient was too expensive and the second one was too bitter.