European protein software startup Cradle launches with backing from Lyft, Twist CEOs

2022-11-18

蛋白降解靶向嵌合体

On his LinkedIn page, Stef van Grieken describes himself as the “purveyor of fine protein.” Van Grieken heads Cradle, a new startup out of the Netherlands and Switzerland that hopes to use machine learning to make it easier for scientists to design new proteins.

Thursday, Cradle launched out of stealth with $5.4 million in seed funding from Kindred Capital and tech fund Index Ventures, alongside a number of notable names — Lyft’s co-founder and president John Zimmer, former DSM CEO Feike Sijbesma, University of California Berkeley professor Patrick Hsu and Twist Bioscience CEO Emily Leproust.

The software biotech houses a platform that marries machine learning and protein design. Scientists can feed the software data on the proteins they want to design, and then use the technology to help engineer the protein sequences with properties that they want, such as making them more stable at higher temperatures or increasing the amount they are expressed.

Cradle is entering a hot field. In August, AI lab DeepMind, owned by the same parent company as Google, showed that it can predict nearly every protein known to man, all 200 million of them. Previously, scientists would go through years of intensive lab work, X-raying and studying them through powerful microscopes, just to learn the structures of proteins. Designing them was even more laborious.

There’s also Minkyung Baek and David Baker’s RoseTTAFold, reverse engineered from the few details DeepMind shared on its software and made publicly available in 2021. (DeepMind made their code available three days later.) And DeepMind exec Mehdi Ghissassi is also an angel investor in Cradle.

Scientists are finally learning how to design proteins from scratch. Drug development may never be the same

Off of his earlier platform Rosetta, David Baker co-founded Cyrus Biotechnology, which gave the protein design software a more user-friendly design. In August, Baker also co-founded Vilya with ARCH Ventures to use machine learning to design therapies — though they did not say for what — out of macrocycles, chemical rings that are smaller than proteins. While earlier methods would have taken extensive leg work to find just a few macrocycles that can get through cells, Baker’s lab used machine learning to predict 84 structures that could get into cells, according to a paper in Cell.

Like Cyrus, Cradle tackles the usability of such platforms. Most scientists aren’t Baker: They don’t have machine learning backgrounds and can’t dedicate themselves to learning the complex code behind Rosetta. Van Grieken told Endpoints News that before starting Cradle, he interviewed some 100 protein engineers, and found that many didn’t have easy access to these machine learning-based platforms. He noted that while there are companies out there that have AI protein platforms, “they have entire teams to sort of run it for themselves.”

Van Grieken’s background is unconventional in the biotech space. In college, he studied philosophy and industrial engineering, but pivoted to working on AI at Google. There, he became interested in applying AI to biology and began hatching the idea of Cradle.

Cradle is still in its early stages. It’s currently partnered with a biotech to improve its platform, though van Grieken declined to disclose the biotech’s identity, saying only that “it’s working on an enzyme.”

“It’s also a little bit novel for some of these scientists to use [Cradle] because they’re used to using Rosetta or they might be using homology,” he said, referring to another type of computer protein modeling, “and so it’s really a process of trying to figure out, how do you make this super user friendly for the scientists?”

Van Grieken said Cradle would be looking to add around five to 10 partners in the next year and make its software free for academics. He said the company’s immediate focus was to “mature the product first,” noting that it doesn’t have the bandwidth to house a large influx of new users.

“There is unfortunately going to be a waitlist,” he said.