Last year, Salesforce, best known for its cloud sales support software (and Slack), led a project called ProGen that uses generative AI to design proteins. If his research moonshot, ProGen, comes to market, it could help find treatments that are more cost-effective than traditional methods, the researchers behind it say in his 2023 research moonshot. I made this claim in a blog post in January.
ProGen culminated with the publication of research results in the journal Nature Biotech showing that AI can successfully create 3D structures for artificial proteins. But outside of the paper, the project didn't have much significance at Salesforce or anywhere else, at least not in a commercial sense.
That is, until recently.
Ali Madani, one of the researchers responsible for ProGen, started a company called Profluent, hoping to bring similar protein production technology out of the lab and into the hands of pharmaceutical companies. In an interview with TechCrunch, Madani described Profluent's mission as “reversing the drug development paradigm,” starting with the patient and treatment needs and working backwards to create “custom-fit” treatment solutions.
“Many drugs, such as enzymes and antibodies, are composed of proteins,” Madani says. “So this is ultimately for patients who will receive AI-designed proteins as medicine.”
While in research at Salesforce, Madani found himself drawn to the similarities between natural languages (like English) and protein “languages.” Madani discovered that proteins – chains of linked amino acids that the body uses for a variety of purposes, from producing hormones to repairing bone and muscle tissue – can be treated like words in a paragraph. Data about proteins can be fed into generative AI models and used to predict entirely new proteins with new functions.
Madani and co-founder Alexander Meeske, an assistant professor of microbiology at the University of Washington, along with Profluent, aim to take this concept a step further by applying it to gene editing.
“Many genetic diseases cannot be solved by: [proteins or enzymes] It is taken directly from nature,” Madani said. “Furthermore, gene editing systems that combine novel features have functional trade-offs that severely limit their reach. In contrast, Profluent can optimize multiple attributes simultaneously to create custom designs. can be achieved. [gene] We provide the best editor for each patient. ”
Not out of left field. Other companies and research groups have demonstrated viable ways in which generative AI can be used to predict proteins.
In 2022, Nvidia released MegaMolBART, a generative AI model trained on a dataset of millions of molecules to search for potential drug targets and predict chemical reactions. Meta trained his model, called ESM-2, on protein sequences. The company claimed that this approach allowed him to predict the sequences of more than 600 million proteins in just two weeks. Also, his AI research lab at Google, DeepMind, has a system called AlphaFold that predicts complete protein structures, achieving speed and accuracy that far exceeds previous, less complex algorithmic methods.
Profluent trains AI models on large datasets (one containing over 40 billion protein sequences) to create new systems as well as fine-tune existing gene editing and protein production systems . Rather than developing treatments on its own, the startup plans to work with external partners to create “genetic medicines” with the most promising paths to approval.
Madani argues that this approach has the potential to significantly reduce the time and money typically required to develop treatments. According to industry group PhRMA, it takes an average of 10 to 15 years to develop a new drug, from initial discovery to regulatory approval. Meanwhile, recent estimates put the cost of developing new drugs at between hundreds of millions of dollars and $2.8 billion.
“Many influential medicines were actually discovered by chance rather than intentionally designed,” Madani said. “[Profluent’s] This ability offers humanity the opportunity to move the most needed solutions in biology from accidental discovery to deliberate design. ”
Berkeley-based Profluent, which has 20 employees, is backed by leading venture capital firms including Spark Capital (which led the company's recent $35 million funding round), Insight Partners, Air Street Capital, AIX Ventures, and Convergent Ventures. Supported. Google Chief Scientist Jeff Dean also contributed, making the platform even more reliable.
Madani said Profluent's focus in the coming months will be on upgrading its AI models, including expanding its training dataset and acquiring customers and partners. You have to be proactive. Rival companies like EevolutionaryScale and Basecamp Research are rapidly training their own protein production models and raising massive VC funding.
“We developed the first platform and demonstrated scientific progress in gene editing,” Madani said. “Now is the time to scale and work with our partners to start delivering solutions that align with our ambitions for the future.”