AI is increasingly being applied to protein design, the process of creating new proteins with specific, target characteristics. Protein design's applications are myriad, but it's a promising way of discovering drug-based treatments to combat diseases and creating new homecare, agriculture, food-based and materials products.

One among the many vendors developing AI tech to design proteins, Arzeda, was founded by a group of researchers at the University of Washington's Baker Lab in 2009. After bootstrapping the company for nine years, Arzeda's founders -- Alexandre Zanghellini, Daniela Grabs, Eric Althoff and David Baker -- decided to bring the Seattle-based firm's protein design platform to market and take on outside VC investment.

"It is difficult to go from lab to commercial scale in industrial biotech, and many companies have failed to do so," Zanghellini, Arzeda's CEO, told TechCrunch. "It is a testament to our progress that we have been able to raise capital in a difficult fundraising environment, with many companies in the space going out of business."

Zanghellini says that Arzeda's tech combines "biophysics-informed" AI models with generative AI methods including large language and diffusion models. To train these models, Arzeda uses a proprietary data set of protein sequences and structures that the startup compiled itself.

Arzeda focuses not on biomedical applications, but on redesigning chemical-heavy goods to use alternative, ostensibly more sustainable AI-designed proteins -- and enzymes. The startup's first product is a natural stevia-based sweetener for an unnamed consumer brand; others in the pipeline include laundry detergents for Unilever and biodegradable materials co-developed with W. L. Gore.

Arzeda also handles the validation and manufacturing of the proteins and enzymes it designs, making money through both sales of its proteins and enzymes and sales of end products that the company creates with contracted partners.

"We have developed filters based on our understanding of structural biology and biophysics to further select designs before they move to experimental testing," Zanghellini said. "All our designs are assayed experimentally, which allows us to weed out hallucinations."

Arzeda's has proven to be a successful strategy. Zanghellini tells me that the 70-person company is generating revenue and that, in addition to Unilever and Gore, Arzeda has clients in AAK (a confectionery fats company) and the Department of Defense. He declined to say more about the Defense Department deal, save that it's "shown very promising results" and "gained a lot of interest."

The growth helped Arzeda close a $38 million oversubscribed funding round led by Sofinnova Partners with participation from Fall Line Capital, Sucden Ventures, Silver Blue (Gore's corporate venture arm), Continental Grain Company, Bunge Ventures and Lewis & Clark Agrifood. Zanghellini said that the proceeds, which brought Arzeda's total raised to $83 million, are being put toward scaling up production of Arzeda's artificial sweetener and supporting the commercialization of other enzymes in the firm's portfolio.

"Our syndicate brings marquee investors covering industrial and agrifood biotech with the domain expertise, network and capital necessary to support Arzeda’s growth," Zanghellini said. "The combination of this latest equity round with our revenues traction provide a comfortable runway and a clear path to 'corporate-EBITDA-positive.'"

This article originally appeared on TechCrunch at https://techcrunch.com/2024/09/17/arzeda-is-using-ai-to-design-proteins-for-artificial-sweeteners-and-more/