Single-cell RNA sequencing has transformed how scientists study biology by letting them measure gene activity in individual cells. This level of detail makes it possible to see how cells differ from one another, but it also introduces a challenge. Inside a single cell, many biological processes happen at the same time. Cell type identity, cell cycle stage, differentiation state, and responses to stress or infection can all overlap, making the data hard to interpret.

Researchers at the Broad Institute of MIT and Harvard developed a new computational approach called CellUntangler to address this problem. Instead of forcing the data into a single explanation, CellUntangler is designed to separate, or disentangle, multiple biological signals that coexist within the same cell.

CellUntangler uses a deep generative model that places cells into a mathematical space made up of several smaller subspaces. Each subspace is tuned to capture a different kind of biological signal. For example, one subspace can represent the cell cycle, while another captures cell type identity. This flexible design allows the method to model complex data more realistically than approaches that focus on just one process.

When the researchers applied CellUntangler to single-cell datasets, they showed that it could successfully separate cell cycle effects from other features like cell type. Importantly, the method also worked for additional signals, including spatial information, tissue dissociation effects, interferon responses, and broader cell identity programs.

One powerful aspect of CellUntangler is that it allows scientists to selectively enhance or remove specific signals at the gene expression level. This means researchers can focus on the biology they care about, such as differentiation or immune activation, without interference from unrelated processes like the cell cycle.

By offering a flexible way to untangle overlapping biological programs, CellUntangler provides a valuable new tool for making sense of complex single-cell RNA sequencing data and extracting clearer biological insights.

Availability – All original code has been deposited in GitHub (https://github.com/Ding-Group/CellUntangler)