AI screening tool identifies hundreds of plant proteins for sustainable emulsifier development

UK-based researchers have developed an AI-powered screening tool that could speed up the search for plant proteins with emulsifying properties. The technology could help food manufacturers reduce the time spent on lengthy trial-and-error testing. 

By combining machine learning and statistical physics, the platform identifies promising protein ingredients that could support the development of more sustainable solutions for applications such as dairy alternatives and plant-based meat. 

The computational screening system has identified nearly 800 plant protein candidates with potential emulsifying properties. Rather than creating finished ingredients, the tool helps scientists identify proteins worth testing before investing extended durations in extraction and laboratory experiments. 

The study, published in Communications Chemistry, was led by Dr. Simha Sridharan, a postdoctoral researcher at the University of Leeds, under the supervision of professor Anwesha Sarkar from the University’s Sarkar Lab. The research also involved AI specialists from the University of Leeds’ School of Food Science and Nutrition and Dr. Rik Sarkar, a machine learning expert at the University of Edinburgh.

The authors take Food Ingredients First behind the scenes to understand how AI is enhancing plant protein discovery and helping F&B manufacturers identify sustainable alternatives to conventional emulsifiers.

Plant protein discovery bottlenecks 

Amid a surge in demand for sustainable and clean label proteins, the plant-based category is moving “from imitation to nutrition,” Innova Market Insights’ analysis highlights.

However, Anwesha Sarkar notes that manufacturers still face “steep innovation challenges to unlock nature’s potential and identify new plant-based ingredients.”Anwesha Sarkar: Extracting, purifying, and testing a single protein for emulsion stability can take months without AI.

“Although the plant kingdom holds millions of proteins capable of acting as functional emulsifiers, the food industry has historically been bottlenecked by slow, resource-intensive, trial-and-error wet laboratory experiments,” she explains.

She argues that until now, replacing conventional emulsifiers, such as monoglycerides or animal proteins (like dairy caseins or egg whites), with plant protein alternatives required scientists to “manually isolate and test candidates one by one using interfacial and colloidal techniques.”

“Because extracting, purifying, and testing a single protein for emulsion stability can take months, scaling this process across thousands of candidates creates an impossible timeline for traditional R&D departments.”

When combined with expert knowledge and experimental validation, the AI tool can “significantly accelerate the discovery of sustainable proteins,” Simha Sridharan tells us.

It also improves the chances of finding ingredients with the desired nutritional and functional properties as well as taste.

AI-powered protein screening 

For the study, Simha Sridharan and Anwesha Sarkar used simulation models and machine learning to understand how proteins interact at oil-water interfaces. This helped the team identify structural features linked to emulsifier performance.

The researchers focused on understanding what makes certain proteins effective emulsifiers at the molecular level. These ingredients often have a characteristic “diblock” structure — two contrasting regions within the protein that allow it to sit at the oil-water interface and stabilize emulsions.Faster protein discovery could support more sustainable and cost-effective F&B innovation.

Rik Sarkar explains that these diblock shapes at the surface of an emulsion droplet “are key to their emulsification action.”

“We developed a mathematical measure called “diblock score” for the similarity of natural proteins to this structure. The AI pipeline divides the proteins into statistically hydrophobic and hydrophilic blocks and looks for protein sequences that are most likely to be diblock emulsifiers.”

Anwesha Sarkar explains that this platform screens “millions of possibilities digitally, separating what will work from what won’t before a scientist ever touches a beaker.”

Commercializing plant proteins 

While the AI model accelerates ingredient discovery, translating promising proteins into commercial ingredients remains the next challenge for the industry.

In F&B applications, the AI model could be particularly valuable where proteins must balance functionality, taste, texture, and cost, according to Simha Sridharan.

“Prime examples include plant-based meat, dairy alternatives, and specialized nutrition products. They must deliver the right texture, taste, functionality, and affordability,” he emphasizes.

An AI-based approach enables researchers to search a much broader range of plant sources and identify proteins with the desired combination of characteristics “far more efficiently than traditional approaches.”Simha Sridharan: AI offers major opportunities in plant-based meat and specialized nutrition, where proteins must meet multiple demands beyond nutrition.

“This can help create products that are more appealing to consumers while also improving sustainability in an efficient and cost-effective manner.”

AI as a tool for food scientists

The AI-based screening offers scientists what Anwesha Sarkar calls “a highly accurate shortlist” of plant proteins.

Scientists can choose those top candidates and evaluate them for emulsion stability, taste, scalability, regulatory, safety, and consumer enjoyment. “It frees up food scientists from tedious screening so they can focus on high-level innovation,” she emphasizes.

Rik Sarkar expects the next generation of plant-based ingredients to come from “searching spaces far larger than any lab could test by hand,” using AI methods such as large language models and physics-informed modeling.

“These approaches will reduce the cost and time required to develop new ingredients and deliver the next generation of plant-based food products.”

“AI will not replace food and ingredient science — it will amplify it,” he concludes.