Nature has done a hell of a job over the last few billion years, but there’s always room for improvement. The proteins in our bodies are great but not always the easiest to produce or package — so Cambrium is looking to design improved molecules that work in similar ways, but can be made at sustainably and at scale, and vegan to boot. And yes, AI is involved.
The company has raised €11 million (around $11.6M) to expand its operations from its proof of concept product, a custom derivative of collagen called NovaColl, to a new slate of structural proteins that could be used in personal care and fashion.
Collagen is a good example of the kind of molecule that makes a lot of sense to make a few changes to. It’s the most common protein in your body, found in practically all your tissues, and drops in collagen levels are associated with wrinkles and other outward signs of aging. The efficacy of collagen supplements, whether oral or topical, is debatable, but it’s definitely both harmless and popular.
There are two problems, though: first, collagen that’s biosimilar to human collagen has to be sourced from animals, generally waste from meat processing. Not great! And second, natural collagen is too large a molecule to really penetrate the skin and do whatever it’s supposed to do.
Cambrium got to work on it in the company’s early stages, isolating the region of the protein that seemed to do the most work and modifying it to be manufacturable in a bioreactor – basically a lot a specialized microbes that eat sugar and excrete the molecule in question. The resulting NovaColl molecule, smaller and easier to make, appears to stimulate collagen production better than the real thing, and it’s done without rendering any organs.
Image Credits: Cambrium
It’s a validation of the approach, showing that there are potentially useful molecules out there being produced in rather nasty ways, and both can be improved.
CEO and founder Mitchell Duffy explained, however, that the company is not interested in medications or enzymes, which are used at industrial scales. Instead, they’re focusing on “structural” proteins, which are valuable for their physical properties.
“A lot of biotech has been focused on enzymes for many years, and structural proteins, until the last decade or so, have fallen by the wayside — but I think they’re some of the higher value proteins out there,” he said. “We’re used to interacting with structural proteins every day, and we need new paradigms in how to create them.”
He suggested silk as an example. Harvested en masse from silkworms, but what we value about it isn’t that it’s bug-based textile. Its value comes from its structural properties, how strands of it are straight and smooth at a molecular level, without the curls or spikes of, say, wool (another valuable structural protein).
“We’re not working on this, but what if you could reduce the pile in some much cheaper bulk material? Then you could have a silk-like feel for a fraction of the price, and it would be vegan, or more sustainable,” he explained.
Targeting structure and function rather than adhering to a strict amino acid sequence (though NovaColl did) confers considerable leeway in the design process.
“Making natural proteins is sometimes more difficult; Evolution optimizes for different things than human needs,” said Duffy. “So if we digitally design proteins, we can optimize for human needs, and you can design for scale. A lot of companies are all about finding the perfect protein and then scaling it up. We’re saying, let’s design a protein that scales to start with.”
Here’s where that AI hook comes in. Surprisingly, it’s not simply algorithmic or bio-specific, because proteins have “that language feel,” as Duffy put it.
“Proteins have a structure like a sentence has a structure, there are words and you can swap them out and get the same meaning but a different intonation or connotation. That’s what we’re doing, we’re forming new sentences. We created a new programing language that lets us put constraints on this generator we have — it’s a model that’s been trained on a ton of data, so it really is a generative AI.”
They’re working on both modifying naturally occurring proteins and designing them from scratch. Rather than taking on something industrial in scale, however, Cambrium is hoping to address high-value, low-volume industries like personal care.
“It’s just a perfect place to start: low minimum quantities, high prices, people are super interested in it, they want to see data but they’re also open to innovation. And textiles treatments are everywhere — not only in woven materials, there are tons used in leather. And there may be more sustainable options,” Duffy said.
The seed funding round was led by Essential Capital, along with SNR, Valor Equity Partners, and HOF Capital, as well as Merantix, out of whose AI venture studio in Berlin the company emerged. Duffy noted that Europe (Cambrium is based in Berlin) has good public funding for this kind of work, allowing them to experiment and develop their processes using grants — “That’s one of the ways we have been able to be so capital efficient and get to market, instead of spending VC dollars on stuff in the lower TRL [tech readiness level] scale.”
NovaColl is shipping to customers now, who are using and testing it in their products, so the market is no longer theoretical. Cambrium’s next molecule or molecules aren’t being revealed yet, but we should hear more soon now that this funding is charging up operations.
