Profluent, an Emeryville, California-based startup, has raised $106 million in venture funding led by Jeff Bezos’s Bezos Expeditions and Altimeter Capital, bringing its total investment to $150 million. The company is pioneering the use of artificial intelligence to design entirely new proteins for drug development and agricultural applications – a field that holds the promise of creating more effective therapies and resilient crops.
The Rise of AI in Biology
The idea behind Profluent emerged from research conducted in 2020, before the widespread availability of tools like ChatGPT. Founder Ali Madani, a machine learning scientist, realized that the same AI architectures used for human language could also be applied to “biological languages” like proteins. Proteins are complex molecules at the core of modern treatments like gene therapies, and represent an advancement over traditional small-molecule drugs. Madani’s work on ProGen at Salesforce demonstrated the possibility of using generative AI to design novel proteins.
How Profluent Works
Profluent’s AI models allow scientists to input desired protein characteristics (such as stability or ease of manufacturing) in plain language, and then generate the corresponding DNA sequence for creation. This moves beyond simply finding existing proteins – the standard approach in drug discovery – and enables custom-design for specific patient needs. The company has already created a database of 115 billion unique proteins, which it calls Protein Atlas, the largest such resource in the world.
The Stakes Are High
The failure rate of new drugs is around 90%, and development costs can reach billions of dollars. This has driven increased interest in AI-powered protein design, though success has proven elusive for many. Profluent isn’t alone in this space: competitors include Isomorphic Labs (a DeepMind spinoff) and Xaira Therapeutics, which raised $1 billion last year.
Scaling Laws and Future Potential
Profluent’s approach relies on “scaling laws” – the principle that more data and computing power lead to better models. The company has demonstrated this principle in protein design, introducing a new foundation model called Profluent E-1 that incorporates evolutionary context.
“One of the reasons Jeff [Bezos] was interested is that we have discovered scaling laws apply to biology. As you gain more and more data, the models get better and better.”
— Ali Madani, Profluent founder
The company’s commercial partners include Revvity, Corteva Agrisciences, and Ensoma, indicating broad interest in its technology. Madani compares the current state of AI-enabled biology to the early days of the internet, suggesting that a fully programmable biological system could create a continuous stream of groundbreaking solutions.
The future of drug discovery and agricultural innovation may hinge on the ability to harness AI’s potential in these complex fields. The current investment signals a growing belief that this is not only possible but inevitable.
