After serving as Pfizer’s top lawyer and now chairing Alnylam’s board, Amy Schulman’s latest biotech bet could be seen as somewhat of a surprise.
The managing partner of venture capital firm Polaris Partners led a $40 million Series A round for Noetik, a Bay Area startup developing artificial intelligence models of cancer. Polaris is typically seen as a more traditional biotech VC, backing companies like Alnylam, Editas Medicine and Alector. But Schulman said Noetik stood out from other AI-focused startups for its distinct approach to using the technology to understand cancer biology and make better medicines.
“There are plenty of smart people, but there isn’t always intellectual chemistry in terms of approach and vision,” Schulman said in an interview. “That really was here.”
Amy SchulmanNoetik was founded in 2022 by ex-Recursion executives, coming out of stealth last year with a $14 million seed round led by DCVC. The AI bio field is dominated by excitement over making models to design better drugs, taking on tasks like optimizing molecules or creating de novo antibodies. Noetik is instead focusing on clinical translation and figuring out which patients would benefit from a particular drug.
“AI companies generally have not thought deeply about that problem,” Noetik CEO Ron Alfa said in an interview. “In my experience, you end up with a really big challenge in understanding which patients should enroll in a trial.”
That’s left cancer R&D with tons of promising new drugs, but less knowledge about who will benefit from them the most. Immunotherapies work for about 12% of cancer patients, but it’s difficult to know which will be responders. Noetik is hoping that by digging deeper into the biology, and with an assist from AI, it can find better ways of bucketing patients and ultimately boost success rates.
So far, the 22-employee company has studied about 1,000 samples of tumors from non-small cell lung cancer patients. Using a variety of cutting-edge techniques, they then study the proteins, RNA and DNA of the cells in those samples. Those data are fed into a self-supervised transformer model, called OCTO.
Alfa is holding back on sharing specifics of what OCTO can do — at least for now. But in general terms, he said the model has shown it can draw inferences across the different ways of studying tumor samples. For instance, it can take staining and protein data from a cancer sample and infer genomic alterations, he said. The company released a technical report last month on the model, although Alfa said the work in that post is already six to eight months old.
“Internally, what we are seeing with the current iterations of our model training is incredibly exciting,” he said.
With the Series A money, Alfa said the company plans to scale up the model, expanding to study other types of cancers, such as sarcomas, ovarian cancer and colorectal cancer, and spend more on compute power. It will also start building a pipeline.
Schulman will join Noetik’s board, which includes Alfa, Noetik CSO Jacob Rinaldi and James Hardiman, a general partner at DCVC. Other new investors in the Series A include Khosla Ventures, Wittington Ventures and Breakout Ventures.