Artificial intelligence (AI)-powered cell capture startup Deepcell Inc. scooped up $20 million in a series A round led by Bow Capital. The funds are earmarked for developing the company’s microfluidics-based technology, building out a cell morphology atlas of more than 400 million cells and advancing a hypothesis-free approach to cell classification and sorting.

Also participating in the round was Andreesen Horowitz, which led Deepcell’s $5 million seed round. With this latest financing, the company has raised a total of $25 million to date.

A 2017 Stanford University spinout, Deepcell employs deep learning and big data to classify and isolate individual cells in a sample to drive precision medicine. The technology melds cutting-edge AI, cell capture and single-cell analysis to sort cells according to highly detailed visual characteristics, providing new insights on cell biology.

Wide range of applications

By using the platform, scientists can keep cells viable for downstream single-cell analysis and can isolate any cell down to frequencies as low as one in a billion, allowing them to retrieve rare and atypical cell states to further precision medicine research.

Maddison Masaeli, CEO and co-founder of Deepcell

“Cell morphology is the bedrock for many clinical applications, and we believe the technology has potential across clinical as well as basic, translational research and biopharma,” Maddison Masaeli, Deepcell co-founder and CEO, told BioWorld. “As we develop and roll out the platform, we plan to work with research partners to show the depth and breadth of the applications before opening it up to the clinical world.

According to Masaeli, the Mountain View, Calif.-based company has built several instruments that are already being used to routinely generate data, supported by a scalable infrastructure around data and Deepcell’s machine learning technology. The company will present its first results next week at the American Society for Cell Biology virtual meeting, and is continuing to expand the number of applications its platform can enable. Current applications include immunotyping, tumor microenvironment and cytology.

The company is also continuing to expand its cell atlas. The database currently contains 400 million cells and nearly a billion images.

“Deepcell’s technology enables users to analyze cells without specifying the markers of interest. This is valuable, because it enables users to analyze cells faster, easier and more flexibly,” Masaeli said. “Importantly, we believe it will usher in a new era of discoveries in cell function and classification.”

With available cell analysis and sorting technologies, users are often forced to guess at various levels of analysis, limiting the quality of their hypotheses, she explained. At the sample level, that could mean hypothesizing the presence of a certain cell in order to sort if the cell has been analyzed. At the cell level, users may need to specify markers of interest before the analysis is completed.

“The way our models are trained allows us to focus on the data rather than a prior hypothesis. In particular, unsupervised learning enables us to deconvolve cell biology similar to the way whole genome sequencing ushered a new era in genomics by enabling the analysis of the genome without the need of hypothesizing what were the regions of interest a priori.” Deepcell holds that same potential for cell biology, Masaeli said.

Higher resolution

Deepcell also differentiates its platform by targeting whole cells rather than cell-free DNA, which enables users to access very precise information about individual cells, including the cell’s full DNA, RNA, epigenetics and proteins.

“Cell-free DNA analysis provides information that is the ‘bulk’ or ‘average’ read out of a sample. It may also be limited by the ways in which circulating DNA is released and degraded after release (e.g., cell death),” Masaeli said. “Single cell analysis allows for a higher resolution view that will allow deeper understanding of cell biology both in tissue and for circulating cells.”

“From its early days in my lab to its launch as a startup, the Deepcell technology has offered the exciting potential of characterizing, identifying and sorting cells without perturbation,” said Euan Ashley, a Stanford professor and co-founder of Deepcell. “Identifying and isolating cells on a spectrum, all the way down to ultra rare, harbors unprecedented potential for understanding single-cell biology and for advancing precision medicine.”

Masaeli agrees. “Cell morphology is a phenotype with a long history in clinical application that has to date been based on the eyes of a human expert. Deepcell is bringing this phenotype into modern use by adding scale, interpretability and actionability, thanks to our innovations in AI, microfluidics and multiomics.”

Deepcell believes its technology has promise in a number of therapeutic categories. Oncology is a major area of interest, with its potential to improve diagnostics and advance precisely targeted therapies. Other areas include immunology, prenatal medicine and cell therapy. Masaeli said the company is looking to partner with precision medicine companies to use its platform in developing new products.