Stately Bio, Inc.

There are always trade-offs to make when studying cells. A biologist either kills the cell to scrutinize its insides or has to run more complex experiments, using tools like fluorescent tags to detect certain biomolecules. A biotech startup believes it has a new solution to this longstanding problem. Stately Bio has raised a $12 million seed round led by AIX Ventures to advance machine learning techniques that recognize signals from simple images of living cells, CEO Frank Li told Endpoints News. The Palo Alto, CA-based startup will officially launch on Friday after three years in stealth mode. Like many in AI bio, Li comes from the world of technology, not biology. He worked as a software engineer at Palantir and Google X before being recruited by Daphne Koller to join her new machine learning team at Calico Life Sciences, an Alphabet-backed biotech. Li spent four years leading Calico’s ML group after Koller left shortly after his arrival to found insitro , another AI-focused biotech. Those were formative years, according to Li, now 34. He reported directly to Calico CEO Art Levinson, the legendary biotech figure who previously ran Genentech, and started to see the potential for ML techniques in cell imaging. In 2019, Li entered a competition at NeurIPS, a top machine learning conference, to find signals from cell images. His team went in thinking 50% accuracy would be a great outcome. They achieved over 99% accuracy, finishing second out of over 800 entrants. What he discovered is the same computer vision techniques that have fueled progress in fields like self-driving cars could also work in biology. “There was a surprising amount of richness to those images that lay beyond our human eye’s ability to comprehend,” Li said. “It was one of those moments that made me go, ‘There’s something really special here.’” Li left Calico in 2021 to found Stately Bio, which began operating in March 2022 with a pre-seed round to focus on cell imaging. Today, Stately is using what Li called “grade-school microscopes” to perform brightfield imaging, which shines light on living cells and takes an image of the other side. Machine learning algorithms find the correlations between cell shapes and textures and their biological functions. “Historically, so much of biology has been learned by looking through a microscope,” Li said. “But it’s only until recently that we’ve been able to take a more quantitative and systematic approach.” Stately isn’t completely alone with this idea. Two years ago, Recursion switched from a more complicated cell-painting technique to brightfield imaging. Other startups like Eikon Therapeutics are working on complementary approaches. Eikon created its own super-resolution microscopes to track single particles inside of living cells. Stately takes a zoomed-out view compared to Eikon, Li said. While observing cells at lower resolutions, the company’s method can watch way more cells, Li said, which is a benefit in building the datasets that power algorithms. For results, Li said Stately improved the ability to differentiate stem cells into liver cells. They have seen 3x to 10x better scores on tests of key liver functions than today’s best techniques, Li said. While the company isn’t yet sharing details about its own pipeline, Li said he sees potential in using lab-made liver cells as a cell replacement therapy for certain liver diseases. Stately is also building AI models that can predict an experiment’s chances of success at very early stages. For instance, the company can take data from the second day of an experiment to predict how those cells will look in a couple of weeks, suggesting if that experiment is worth carrying out or stopping early.