Last year, Salesforce, renowned for its cloud sales software (and Slack), initiated an ambitious project called ProGen aimed at designing proteins using generative AI. This innovative research effort, detailed in a January 2023 blog post, suggested that ProGen could revolutionize the discovery of medical treatments, making the process more efficient and cost-effective compared to conventional methods.
ProGen’s research was later published in the journal Nature Biotech, demonstrating how AI could successfully generate the 3D structures of artificial proteins. However, until now, the initiative didn’t translate into considerable commercial impact for Salesforce or other sectors.
Recently, everything changed. One of ProGen's lead researchers, Ali Madani, has founded a company called Profluent. His vision is to leverage similar protein-generating technology and make it accessible to pharmaceutical companies. In an interview, Madani articulated Profluent's goal of “reversing the drug development paradigm.” The company aims to focus on patient and therapeutic needs and innovate “custom-fit” treatment solutions.
“Many drugs – including enzymes and antibodies – are made of proteins,” Madani explained. “Ultimately, this is aimed at patients who could benefit from AI-designed proteins as treatment options.”
While at Salesforce, Madani recognized fascinating parallels between natural language (like English) and the “language” of proteins. Proteins, which are chains of amino acids utilized by the body for various functions—such as hormone production and tissue repair—can be viewed as akin to words in a sentence. When input into a generative AI model, data about proteins can predict entirely new proteins with unique functions.
Working alongside co-founder Alexander Meeske, an assistant professor of microbiology at the University of Washington, Profluent aims to advance these concepts by applying them to gene editing technologies.
“Many genetic diseases can’t be resolved with proteins or enzymes directly sourced from nature,” Madani noted. “Additionally, existing gene editing systems, while innovative, often experience functional trade-offs that restrict their versatility. Profluent offers the ability to optimize several characteristics simultaneously, creating custom-designed gene editors tailored to individual patients.”
This focus isn’t entirely novel; many companies and research groups have successfully employed generative AI to predict protein structures. For instance, in 2022, Nvidia introduced a generative AI model named MegaMolBART, trained on millions of molecules to seek drug targets and predict chemical reactions. Similarly, Meta developed a model called ESM-2 that could forecast sequences for over 600 million proteins within just two weeks. DeepMind has advanced this field with AlphaFold, giving unprecedented speed and accuracy in predicting complete protein structures compared to traditional algorithms.
Profluent is harnessing AI models trained on extensive datasets—containing over 40 billion protein sequences—to innovate new and refine existing gene-editing and protein production systems. Instead of developing medicines independently, the startup plans to collaborate with external partners to create “genetic medicines” that show the most promise for market approval.
Madani believes this strategy could significantly reduce the time and budget usually necessary to bring a new treatment to market. Industry reports by PhRMA indicate that it typically takes 10-15 years to bring a new medicine from initial discovery to regulatory approval, with development costs often soaring between hundreds of millions to $2.8 billion.
“Many significant medicines have been discovered by accident rather than through deliberate design,” Madani stated. “Profluent’s capabilities allow us to shift from serendipitous discoveries to intentional design of essential biological solutions.”
Based in Berkeley, Profluent employs approximately 20 people and is backed by influential venture capital firms, including Spark Capital, which led a recent $35 million funding round, along with Insight Partners, Air Street Capital, AIX Ventures, and Convergent Ventures. Google's chief scientist, Jeff Dean, has also supported the venture, underscoring its credibility.
In the coming months, Profluent plans to enhance its AI models by expanding its training datasets, according to Madani, as well as ramping up customer and partner acquisition efforts. Competition is heating up, with rivals like EvolutionaryScale and Basecamp Research quickly refining their own protein-generating models and securing substantial venture financing.
“We’ve built our foundational platform and achieved significant scientific advances in gene editing,” Madani emphasized. “Now is the critical moment to scale our operations and partner with organizations that share our vision for the future.”
Ranking
