Scientists at the UK’s National Alternative Protein Innovation Centre have used artificial intelligence to identify nearly 800 plant proteins that could potentially be used as emulsifiers in food and personal care products.
The research team at the University of Leeds combined artificial intelligence with statistical physics to predict which proteins could perform this function. Many of the nearly 800 proteins identified had not previously been considered as potential emulsifiers.
Emulsifiers allow oil and water to mix and help keep products stable. They are used in foods such as ice cream, sauces and mayonnaise, as well as cosmetics, pharmaceuticals and other industrial products.
Many existing emulsifiers are derived from animal proteins, including whey, casein and eggs. Researchers are increasingly looking for natural and more sustainable alternatives, but conventional testing presents a major challenge because millions of plant proteins have potentially useful functional properties.
The Leeds researchers first used a simulation model based on statistical physics to study how proteins interact at the boundary between oil and water. They then used machine learning to identify the sections and characteristics of proteins that influence this behaviour.
The approach allowed them to narrow millions of possible proteins down to candidates worth testing in the laboratory. Several commercially available proteins were subsequently tested, with pea and potato proteins showing effective emulsifying properties that supported the model’s predictions.
Researchers say the method could reduce years of trial-and-error testing and speed up early-stage ingredient development. The work also brings together food science, protein chemistry, statistical physics and AI to address challenges in developing sustainable proteins.
