How AI is helping to change byproducts to products people buy

The world throws away a significant amount of food every day. But what if we could turn some of that waste into tasty, nutritious protein to feed people and reduce emissions in the process?

Researchers at Leeds have teamed up with business and partners in Australia to use AI-supported fermentation in an attempt to do just that.

You might think the steady march of time and our human desire to improve our lot are slowly but surely making famine a thing of the past.

Think again.

According to the World Health Organization, almost 30 per cent of people on the planet did not have constant access to food in 2022, with around 900 million of those 2.4 billion facing “severe food insecurity”.

Yet this picture doesn’t exist because we don’t have enough food.

Producing profitable proteins

Making meat-free food can be an expensive thing to do.

“At present, when companies want to produce lots of alternative proteins, they often have to rely on the typical biomass sources that are commonly available." Alex Bowler, Research Assistant in the School of Food Science and Nutrition said.

“These are usually high in sugar and quite costly, so finding new sources that we can use in the fermentation process offers the chance to make things more affordable.”

The substrates in question – essentially surfaces that organisms grow on to create the desired protein or nutrients – are often glucose and sucrose, and can be responsible for the high production cost of things like plant-based burgers and steaks.

Replacing them with substrates of cheaper products could not only cut costs for manufacturers already investing in alternative proteins, but encourage more in the industry to turn sidestreams into products that could be as lucrative as they are ethical.

As the Global Food Institute describes it, “using sidestreams as ingredients in production will improve the sustainability of alternative protein manufacturing as we move closer to a sustainable, circular bioeconomy”.

Alex is part of a team led by Professor of Artificial Intelligence in Food Nik Watson, as is experienced Teaching Fellow Dr Darren Greetham.

“I’m essentially a yeast microbiologist,” Darren explained. “I’ve spent the last 20 years looking at how yeast overcomes stress, so that’s what I bring to this project.”

Tapping into efficient tools

Together with the Commonwealth Scientific and Industrial Research Organisation (CSIRO) and several Australian businesses, Darren, Alex and the team at Leeds are using a range of different skill sets to tackle some of the challenges of alternative protein production.

The collaboration is combining the data collection and model development capabilities available at Leeds with the CSIRO’s experience of scaling up processes to deliver a real-world impact.

In trying to extract proteins through fermentation, the possibilities are endless.

That doesn’t mean it’s a simple case of throwing things into the mix and seeing what sticks, which could be a time-consuming process of expensive, laborious trial and error.

Instead, Alex takes advantage of the power of machine learning to accurately and cost-effectively predict how the many fermentation factors at play, such as temperature and pH level, would all interact.

“It involves going back through our lab data and extensive external literature to collect as much useful data as possible and feed that into the AI,” Alex explained.

There is, however, a degree of human serendipity involved in the project, too.

Collaborative discoveries

“A number of years ago before I joined Leeds, I went to a conference and a big, tall farmer came up and asked me whether I was interested in Jerusalem artichokes. To be honest, I wasn’t initially – but his passion for them was infectious.

“He grew them on a commercial farm, and about a week later he emailed me saying that he was in the area and wanted to bring me a pint of Jerusalem artichoke juice to ferment, which we did on quite a large scale.

“The yeast absolutely loved this fructose-rich, sugary black liquid, which he had boiled on a kitchen hob for about two weeks.”

Using hydrolysis, Darren streamlined this slow process so that the same volume of Jerusalem artichoke liquid – which is full of good-for-the-gut inulin – could be made in just three hours.

“Fast-forward to me moving to Leeds, and Nik asked me if I knew anyone with a vegetable substrate that we could use in this project with our partners in Australia – so that farmer came to mind right away.”

Preventing sidestream side effects

It’s just one example of how underutilised crops can be used to produce alternative sources of protein, although much of the material used by the team derives from the sidestreams generated by food manufacturers.

The use of AI, meanwhile, is as much about eliminating undesirable characteristics as it is generating nutritious, cost-effective proteins.

“When soft fruits like strawberries and raspberries break down, they make polyphenols and other ketones and aldehydes that can be very unpleasant,” Darren said. “Our machine learning process can help to get rid of these or mitigate them; it’s useful not only in making protein, ethanol, vitamins and so on, but also for stopping other things from creeping in.”

Solving problems across continents

The work of Alex, Darren and the team at Leeds are set to create products with a low environmental impact that contain fewer additives and go through less processing.

The latter could offer a key advantage to brands looking to market their products in the future, given almost half of participants in one survey said they were worried that plant-based meat is “too processed” or “unnatural”.

By partnering with the CSIRO, the research breakthroughs achieved by the team at Leeds can be scaled up to deliver a bigger real-world impact.

Both organisations stand to improve their AI capabilities, producing models that can be used by others around the world to optimise efficient fermentation processes and identify new food side streams.

Ancient practice, meet modern methods

For Darren, it’s an exciting step forward that could lead to other innovations in the future.

We haven’t even thought about things like future foods or functional polymers, so there’s all kinds of things we could explore further by using these techniques as the building blocks.

Humans have taken advantage of fermentation since ancient history, with the earliest archaeological evidence for beer production dating back some 13,000 years.

Now, the research efforts of scientists at Leeds and partners on the other side of the planet are breathing new life into the process with modern technologies – all in the name of reducing waste, famine, malnutrition and food insecurity.