When asked about the focus of his new company, Ron Alfa, MD, PhD, begins with an epistemological explanation, and the definition of disease (he is a trained physician, after all.) In some cancers, he explained, a tumor type is defined by a genotype. This allows for the development of precision therapeutics and the stratification of patients. But these lines have yet to be drawn in immuno-oncology. “We haven’t been able to define patient populations by the underlying immune biology in a clear and simple way,” he told GEN. To support his assertion, he points to the dearth of new immuno-oncology molecules developed over the past decade or two (with the exception of PD-L1) with promising preclinical data.
Why hasn’t there been more success in this area? Alfa cites a lack of understanding of which tumors are most likely to respond to certain mechanisms. Because, he said, we don’t have a system that helps us understand that some tumors are inhibiting the immune system by mechanism A, some by mechanism B, and others by mechanism C. Once that information is uncovered, immuno-oncology can be pushed into precision oncology, and tumors can be classified based on their mechanisms of immune resistance and immune biology.
But immune biology does not cleanly map to genomics. So, Alfa and Jacob Rinaldi, PhD, co-founder and current CSO at Noetik—who met during their graduate school days at Stanford University—have founded their company to solve this problem.
First, they sought to understand what are the best data to 1) understand tumor biology and 2) use in machine learning models. Enter spatial biology.
Not your normal spatial customer
Noetik is not a spatial company, Alfa asserted. But they use a lot of spatial. They are, according to Alfa, the biggest users of two new platforms from two different spatial companies—one for proteomics and the other transcriptomics.
For Noetik, spatial is the right dataset for machine learning to help them achieve their goal, Alfa said. But the spatial going on at Noetik is different from how other researchers are using the new technology. “What we’re doing doesn’t look much like what [academics are] doing.” Because Noetik is using spatial through the lens of building a dataset for machine learning. To that end, they need a lot of data.
“I’ve literally told both of [the spatial companies] that we are not your normal customer.” Academics have a totally different set of requirements than we do. Noetik needs to scale, run the machine 24 hours a day, get the data onto the cloud, have QC metrics, and other industrialization requests.
They have, in six months, generated hundreds of patients worth of data. For each patient, they have genomic data, H&E staining, proteomics, and transcriptomics.
The goal is for the models to learn from that information. They are training the models to do something that humans can’t do: to classify these tumors into tumor immune subsets in ways that are therapeutically important. To do this, they use self-surpervised learning, allowing the models to learn internal representations of the data themselves. Eventually, they hope to learn tumor biology from the models.
The company’s work points in the direction of building a portfolio of molecules that are specifically directed at tumor subsets—though there is no pipeline yet. But they will start from the patient, using reverse translation. Noetik’s platform does not help predict a molecule or design a drug. Hopefully, it uncovers novel biology, and in doing so, informs what drug they should make.
Driven by frustration
Over the course of Ron Alfa’s lifetime, he has pursued things that he finds frustrating. “It’s always been very frustrating that we don’t have better cancer drugs. It’s one of these things that [you think], this is ridiculous! We know so much about biology, why don’t we have better cancer drugs?”
Alfa and Rinaldi left their previous posts at Recursion in February of last year, and Noetik was incorporated in June of the same year—just 15 months ago. They built the company remotely but have the largest cluster in South SF and the second largest group in Boulder. The company of 18 employees announced last week that it closed an oversubscribed $14M seed financing round led by DCVC.
The challenge Noetik faces is the same one facing many biotech and pharma companies: developing better cancer drugs. But Noetik’s approach uses different data to try to solve that problem.
The name Noetik is derived from a Greek word for intellectual. They chose the name for two reasons: one, the relationship between the word intellectual and artificial intelligence. Two, the hope that artificial intelligence is smart enough to understand patient groups, which will lead to better cancer therapeutics.
