Profluent uses AI to find drugs, inspired by Salesforce research and supported by Jeff Dean

Last yr, Salesforce, the corporate best known for its cloud sales enablement software (and Slack), led a project called ProGen to design proteins using generative AI. ProGen is a research project that, if dropped at market, could help uncover medical treatments more cost-effectively than traditional methods, in response to the researchers behind it claims in a blog post from January 2023.

ProGen culminated in a study published within the journal Nature Biotech that showed that AI could successfully generate the 3D structures of artificial proteins. But apart from the newspaper, the project did not have much significance at Salesforce or elsewhere – not less than not in a industrial sense.

That is, until recently.

One of the researchers answerable for ProGen, Ali Madani, founded an organization They will flow, that he hopes to bring similar protein-making technologies from the lab into the hands of pharmaceutical corporations. In an interview with TechCrunch, Madani describes Profluent’s mission as “reversing the drug development paradigm,” starting with patient and therapy needs and dealing backwards to create “tailored” treatment solutions.

“Many drugs – enzymes and antibodies, for instance – are fabricated from proteins,” Madani said. “So ultimately that is for patients who would receive an AI-developed protein as a drug.”

While working within the research department at Salesforce Madani was drawn to the parallels between natural language (e.g. English) and the “language” of proteins. Proteins — chains of interconnected amino acids that the body uses for quite a lot of purposes, from making hormones to repairing bone and muscle tissue — could be treated like words in a paragraph, Madani discovered. When fed right into a generative AI model, data about proteins could be used to predict entirely latest proteins with novel functions.

With Profluent, Madani and co-founder Alexander Meeske, assistant professor of microbiology on the University of Washington, need to take the concept a step further by applying it to gene editing.

“Many genetic diseases can’t be corrected by (proteins or enzymes) directly from nature,” Madani said. “Furthermore, gene editing systems tuned to latest capabilities suffer from functional trade-offs that significantly limit their reach. In contrast, Profluent can optimize multiple attributes concurrently to acquire a tailored (gene) editor that completely suits each patient.”

It’s not out of left field. Other corporations and research groups have demonstrated viable ways to make use of generative AI to predict proteins.

Nvidia released a generative AI model in 2022, MegaMolBART, trained on a dataset of thousands and thousands of molecules to look for potential drug targets and predict chemical reactions. Meta educated a model called ESM-2 for protein sequences, an approach that the corporate said was in a position to predict sequences for greater than 600 million proteins in only two weeks. And DeepMind, Google’s AI research lab, has a system called AlphaFold that predicts complete protein structures at a speed and accuracy that far surpasses older, less complex algorithmic methods.

Profluent trains AI models on massive data sets – data sets containing over 40 billion protein sequences – to create latest gene editing and protein production systems and optimize existing systems. Instead of developing treatments itself, the startup plans to work with external partners to develop “genetic medicines” with essentially the most promising approval pathways.

Madani claims this approach could dramatically reduce the time — and capital — typically required to develop a treatment. According to industry group PhRMA, the event of a brand new drug takes a mean of 10 to fifteen years from initial discovery to regulatory approval. Youngest Estimates The cost of developing a brand new drug now ranges from several hundred million to $2.8 billion.

“Many effective drugs were actually discovered by accident fairly than intentionally developed,” Madani said. “(Profluent’s) capabilities offer humanity the chance to maneuver from accidental discovery to the conscious development of our most needed solutions in biology.”

Berkeley-based Profluent, which has 20 employees, is backed by major enterprise capitalists including Spark Capital (which led the corporate’s recent $35 million funding round), Insight Partners, Air Street Capital, AIX Ventures and Convergent Ventures. Google chief scientist Jeff Dean also contributed and gave the platform additional credibility.

Profluent’s focus in the following few months will probably be on updating its AI models, including by expanding training data sets, says Madani, in addition to customer and partner acquisition. It must move aggressively; Competitors, including EvolutionaryScale and Basecamp Research, are quickly training their very own protein-generating models and raising huge sums of VC money.

“We developed our first platform and made scientific breakthroughs in gene editing,” Madani said. “Now is the time to scale with partners and enable solutions that meet our ambitions for the long run.”