AI identifies 787 proteins with emulsifier potential

Finding a plant protein that can work as an emulsifier sounds easier than it is. The range of options is vast, while laboratory research requires considerable time and resources. Researchers are therefore using artificial intelligence as an initial filter. Their method flagged hundreds of proteins, including many unexpected options.

A targeted search for emulsifiers

Emulsifiers help oil and water blend and remain stable. They are used in products including sauces, ice cream, and mayonnaise. Interest in natural and more sustainable alternatives is growing. Researchers are therefore also seeking replacements for casein and whey.

Scientists at the University of Leeds developed a new approach. The method combines artificial intelligence with statistical physics. A simulation model calculates how proteins behave at the interface between oil and water. Artificial intelligence then determines which protein characteristics influence this behavior. This brings the most promising proteins to the forefront.

787 potential proteins

The researchers initially selected 1,950 plant proteins for further analysis. The simulation model produced usable results for 1,858 proteins. Within this group, they identified 787 proteins with a suitable structure. Many had not previously been considered as potential emulsifiers.

During laboratory testing, the researchers examined two plant proteins, among others. These were legumin B from peas and patatin from potatoes. Both proteins were chemically unfolded beforehand.

The proteins reduced interfacial tension by roughly the same amount as beta-casein. They also formed oil-in-water emulsions with comparable droplet sizes. However, beta-casein produced the narrowest and most uniform droplet size distribution.

This computational method reduces the need for separate laboratory tests. It allows researchers to select proteins more precisely for further development.

Leeds.ac.uk