Unlocking the “Morpholome” With AI-Powered Imaging

Cell morphology is a rich source of biological information that is indicative of phenotype and function.  However, molecular-driven cytologic analysis is limited due to a lack of robust, standardized and low-cost techniques for preparing cytology samples. Molecular analysis is also destructive and subjective.

Technology Networks had the pleasure of speaking with Dr. Maddison Masaeli, co-founder and chief executive officer at Deepcell to learn about the REM-I platform, which is a single-cell analysis tool that utilizes AI-powered imaging to magnify insights into a cell’s phenotype and function. Such information could help catalyze new methods of discovery in cancer biology, developmental biology, stem cell biology, gene therapy and functional screening.

Kate Robinson (KR): What information can cell morphology provide and how clinically relevant is this data?

Maddison Masaeli (MM): Cell morphology has been one of the first and most fundamental phenotypes used to study cells in labs and clinical settings. Identification uses microscopes, a process that’s inherently slow, difficult to scale and relies on human interpretation. As a result, tools for cellular quantification have not developed as quickly as other “omics”-based methods for biological research.

Deepcell seeks to bring scalability to morphology and to unlock the “morpholome” by bringing AI-powered imaging and analysis to cell biology, diagnostics and beyond. Deepcell’s tools make a discovery-based approach to cellular analysis possible.

KR: Can you tell us about the Deepcell platform and its applications?

MM: The Deepcell platform, which includes a benchtop instrument and consumables, is a cellular analysis tool that combines high speed imaging, sorting, high-dimensional AI characterization and analysis software to enable users to image, analyze, store and sort cells for downstream analysis using other biological research methods.

Unlike traditional flow cytometry approaches, Deepcell brings a truly unbiased approach to cell biology research by enabling laboratories to study cells in flow without labels or staining and identify cellular features of interest in a hypothesis-free way. Leveraging the Deepcell platform’s 6-way sorting capabilities, researchers can further sort cells of interest for downstream analysis using custom, morphology-based classifications.

Today, morpholomics offers potential to a range of research applications including cancer research, developmental biology research, cell and gene therapy development and manufacturing and drug and functional screening, but in the future, we envision morpholomics data also being used in a range of diagnostics applications.

KR: How does the REM-I platform enhance typical cell morphology data using artificial intelligence?

MM: The REM-I platform is powered by Deepcell’s artificial intelligence model, the Human Foundation Model (HFM). The HFM extracts visual features from a large and diverse set of cell images in a self-supervised fashion, without prior knowledge of specific cell types, cell preparations or other application-specific markers for an unbounded approach to hypothesis generation and testing. The platform uses deep learning and computer vision to quantify, interpret and scale morphological analysis of single cells, and enables creation of massive morphology-based datasets. The platform not only enables the researcher to review the individual cell images and groupings of cells, but also to sort morphologically desired cells into up to six wells for additional downstream analysis.

KR: Are there any examples of the platform being used that you would like to highlight?

MM: Providing top researchers with early access to our technology has been a crucial initiative for Deepcell. It has allowed us to test and validate the product’s performance in real-world settings. The participants have had the opportunity to collaborate closely with our development team, sharing their expertise and insights. To date, our partners have used the technology for a variety of research areas, including liquid biopsy, drug screening, cancer biology and CRISPR screening. The results of some of these studies are published or are in the process of publication for public access.

Dr. Maddison Masaeli was speaking to Kate Robinson, Assistant Editor for Technology Networks.