Profluent founder Ali Madani
jin han
Profluent has raised $106 million to scale up use of its AI models for biology in drug development and agriculture.
Ali Madani started thinking about how AI could be programmed for biology in 2020, years before the launch of ChatGPT. The machine learning scientist was working at Salesforce, which that year launched a moonshot project, called ProGen, to design novel proteins with generative AI. “The same architecture used for English, you can use for biological languages like proteins,” he told Forbes.
He left Salesforce in 2022 and teamed up with Alexander Meeske, head of a research lab at the University of Washington, to bring that promise to life. Now, his Emeryville, California-based startup Profluent’s AI models enable scientists to explain the properties they want in a protein in human language (like stability or ease of manufacturing), and then output a DNA recipe to create that protein.
Madani, who has a Ph.D. from the University of California, Berkeley and was the lead author of a Nature Biotechnology paper on ProGen, believes focusing on proteins could unlock groundbreaking new drugs. Proteins are large molecules that are significantly more complex than the small molecules that are the basis of many existing drugs, but allow for newer treatments like gene therapies. And he hopes it also will lead to breakthroughs in agriculture, where researchers hope to create more resilient and sustainable crops.
“The proposition of making biology programmable is going to enable blockbuster drugs, and solutions across therapeutics, diagnostics and agriculture—and it’s going to require a lot of capital,” Madani said.
To that end, Profluent said Wednesday that it had raised $106 million in new venture funding led by Jeff Bezos’s Bezos Expeditions and Altimeter Capital, bringing total investment to $150 million. With the new financing, Profluent’s valuation is approaching $1 billion. Its commercial partners include Revvity, an $11 billion (market cap) biotech; Corteva Agrisciences, the agricultural spinoff of DuPont; and VC-backed Ensoma, which is working on treatments for genetic diseases and cancer.
“The proposition of making biology programmable is going to enable blockbuster drugs … and it’s going to require a lot of capital.”
Companies like Recursion have been trying for years to bring AI to drug discovery, though the efforts have proven more difficult than researchers had originally hoped. But with 90% of new drugs failing and the cost to develop new ones running in the billions, an increasing number of companies are working on tackling the problem with AI-enabled protein design. Profluent is up against heavyweights like Isomorphic Labs, the spinoff of Google’s AI research lab DeepMind, and startups like Xaira Therapeutics, which emerged from stealth last year with $1 billion in funding.
Profluent’s goal is not just to use AI to find existing proteins, as is typically the way existing drug development is done, but to custom-design completely new proteins for a patient’s needs.
To date, Profluent has created a database that it calls Protein Atlas, comprising 115 billion unique proteins, which it says is the largest such protein data resource in the world. All that data, combined with more compute power, should help it build bigger, better models — a concept known as “scaling laws.” Earlier this year, Profluent said that it had demonstrated scaling laws work for models that design proteins. Last week, it introduced a new foundation model, called Profluent E-1, that provides evolutionary context.
For the latest healthcare news, sign up for Forbes’ InnovationRx newsletter.
“One of the reasons Jeff [Bezos] was interested is that we have discovered scaling laws” apply to biology, Madani said. “As you gain more and more data, the models get better and better.”
Investor Jamin Ball, a partner at Altimeter Capital, who first met Madani about 18 months ago, and got to know him on walks around a bio conference in San Diego last year, said there’s “a massive, massive, massive opportunity” in the ability of scientists to go from happenstance in drug discovery to bespoke design. “We think the next frontier in AI will be biology and drug discovery,” he said.
“We are still really early,” Madani said, comparing the current state of AI-enabled biology to the early days of the Internet. “If we can have a machine that can truly make biology programmable, we will have a conveyer belt of blockbuster solutions.”
MORE FROM FORBES