Many of the food industry’s most persistent problems exist not in the realm of business and economics – scaling, costs of materials and the like – but in nature itself.
Crops, which the food industry relies on, are vulnerable to heat, drought and disease. Allergens, benign to some consumers, are deadly to others. Forever chemicals have innumerable practical uses in food and agriculture, yet can also seep into food and increase health risks for consumers.
With the combination of protein design and recent developments in AI, all these issues could be solved.
Many of the issues seen here are the way they are because of the way that proteins work – proteins that give chemicals and allergens their properties. New proteins, with different functionalities, can be designed, but until recently, this process was so time-consuming it yielded very piecemeal results.
With AI, this is no longer the case. With AI, progress that once took years can be achieved in months.
What is protein design?
Within nature, proteins fulfil a wide range of functions; regulating photosynthesis in crops, taste and texture in food, and the immune system in human beings.
With the use of computational tools, protein structures can be designed completely from scratch, giving researchers the ability to create entirely new proteins, with entirely new functionalities.
“We are in the business of creating completely new proteins . . . we create them entirely from scratch, built around some intended function,” explains Ian C Haydon, head of communications and AI policy at the Institute of Protein Design at the University of Washington.
While some synthetic proteins are modified versions of existing proteins, those created by the institute are built “atom by atom”, completely from scratch. In other words, not only does the institute create digital models of these proteins, but it can create them physically as well.
The institute uses ‘deep learning,’ which is a method of AI machine learning inspired by the human brain. Until recently, protein design took much longer, but recent developments within AI have sped things up.
Another tool that uses deep learning is AlphaFold, an AI tool that can predict the structure of proteins. The tool, the creators of which won the 2024 Nobel Prize in Chemistry, drastically improves the ability of researchers to predict protein structure. The capabilities of similar advanced AI tools make protein design much more efficient as well.
With such tools, as much can be achieved by an undergraduate student visiting the lab for a month or two than could previously have been in a three-year Masters degree, explains Haydon.
Not only are these tools cutting down on research time, but they are also making projects more ambitious. Proteins that are designed now are “much closer to what you see out in nature” than those before these developments.
How can protein design destroy PFAS?
PFAS, or forever chemicals, are one of the major problems in food and beverage today. Given the moniker ‘forever’ because they remain in the environment for thousands of years without degrading, PFAS are often found in non-stick cookware, food packaging and fertilisers used on crops. Their function is to repel oil and water and resist heat.
However, when they make their way into food itself, they can be harmful to health. Specifically, they have been linked to liver damage, fertility issues, thyroid disease, and increased cancer risk. Despite this, they have such a wide range of potential that they cannot be easily cast aside.
The problem with PFAS is also what makes it so useful in the food industry, Haydon explains: PFASs are “extremely stable”, meaning both that they fulfil their functions well, and that they’re very hard to get rid of.
The reason that they are so stable is linked to the inclusion of halogens, typically the element fluorine. The chemical bond between fluorine and carbon, Haydon explains, is the strongest bond that can be formed with carbon. That’s what gives chemicals such as PFAS their durability.
The Institute’s researchers are aiming to design catalysts that can not only recognise where this bond is, but cut the bond apart.
This will either destroy the compound, turning it into something more benign, or cleave off the chemical feature that creates such stability, which will allow it to break down more easily.
Because fluorine-type compounds are not common in biology, there hasn’t been much pressure from evolution to come up with the type of catalyst that will break them apart. With computational design tools, however, such catalysts can be created.
For dealing with PFAS specifically, the Protein Design Institute has at least 20 active research projects. The software for doing so is also open-source, meaning that many other research teams will be able to access it.
How can protein design improve crop resilience?
Crops all over the world have been impacted by the changing climate. Coffee crops in Brazil have been feeling the pinch of drought. Cocoa crops have been experiencing the scourge of swollen shoot virus. Crops worldwide have seen the effects of rising temperatures push down yields and push up prices as heatwaves wrack the world.
Protein design has the potential to improve resistance in these three areas: heat, drought and disease.
Heatwaves can be “devastating for plant cells,” explains Haydon. “The proteins inside those plant cells are not stable at higher temperatures. They unfold, they fall apart, sometimes irreversibly. So that’s why even a short exposure to high heat can give a generational damage to a plant. Those proteins break, and it delays development, it delays growth.”
In this case, rather than inventing new proteins, the researchers can use these tools to modify existing proteins. Making proteins more heat tolerant is one of the modifications that such technology is able to do well, and researchers have become skilled at doing so.
In fact, they have become so successful that they have seen examples of molecules which would normally fail at 40°C remain intact at 80°C.
Their work is not done, however. The next stage is to take these modifications to plants themselves. They need to work out which proteins are failing first, which cause the most significant effect, and what is the potential to improve stability for each one.
If this works, they next need to work out how to deploy these modified proteins within crops themselves, in a way that will satisfy key stakeholders such as regulators, buyers, and farmers.
“This is a stakeholder and regulatory issue as much as a scientific and technical issue.”
For this project, they are also collaborating with the Innovative Genomics Institute, run by Jennifer Doudna who won the 2020 Nobel Prize in Chemistry for her work in CRISPR gene editing. Computational protein design and gene editing, Haydon says, are a “match made in heaven.”
Drought, on the other hand, has a very complex effect on a plant. The lack of water “can trigger all kinds of crises within a plant. There, their chemistry goes haywire. They try to go on life support. They shut down their metabolism in an attempt to survive,” Haydon explains.
To mitigate this problem, researchers are looking at ways of dealing with oxidative stress, which is part of the plant’s emergency response to not getting enough water. Improving this has the potential to help plants deal with periods of drought better.
Finally, improving pest resistance would involve both giving plants the ability to detect an infection, and equipping them with the molecules allowing them to fight this infection.
How can protein design help people with allergies?
The Food and Agriculture Organisation of the United Nations (FAO) estimates that more than 220 million people worldwide have food allergies. Combating the effects of allergies is thus important for many people.
One particularly successful project in this area, which began as an undergraduate research project, was one combating the effects of coeliac disease, an autoimmune disease triggered by gluten.
The project developed an enzyme which those suffering from coeliac disease could eat with bread in the form of a pill, that could digest gluten on behalf of the person consuming it.
“It’s taking what would otherwise be in a problematic peptide in your food and physically cleaving it so that it can’t produce an immune response or inflammatory response in the intestines.”
This project eventually led to a spin-out company called PvP Biologics, which was then purchased by a large pharmaceutical company, Takeda Pharmaceuticals.
The name ‘PvP’ comes from ‘protein vs. protein’, as the company created an enzyme (a protein) to attack gluten (another protein).
What other applications does protein design have?
Beyond these, there are many other food and agriculture-related applications for this technology that it may be utilised for in the future.
For example, there is the potential to rebuild photosynthesis, so that plants can create fixed nitrogen to replace synthetic fertiliser.
Furthermore, they may be able to design a nutritionally optimal plant-based protein, optimising the nutritional quality of protein from crops.
While the researchers have wanted to do these for years, before this, the tools haven’t been good enough. Now, a whole lot more is possible.
