CellWalker2: A New Approach to Decoding Single-Cell Genomic Data

Biologists often face the challenge of interpreting vast datasets from experiments that profile thousands of individual cells, much like trying to assemble a puzzle without a reference image. To make sense of this data, scientists define cell types, but comparing these types across different studies can be difficult due to varying classification methods.

Researchers at Gladstone Institutes have developed a computational tool called CellWalker2 to address this issue. This method enables scientists to determine relationships between cell types and identify groupings that may influence health. As Katie Pollard, who led the study published in Cell Genomics, explains, “Previous methods did not leverage relationships between cell types. With our new tool, we use the fact that some cell types are siblings rather than distant cousins.”

CellWalker2 uses hierarchical relationships to distinguish between similar cell types, such as immature and mature neurons, by first grouping them broadly and then refining the categories. If the tool cannot identify a precise cell type, it assigns a broader label. Zhirui Hu, first author of the study, notes, “Cell types aren’t random categories. Some are very closely related...while others are fundamentally different. CellWalker2 takes those relationships into account.”

The tool is particularly useful for analyzing data from single-cell ATAC-seq, which reveals accessible regions of DNA that may influence disease. CellWalker2 connects multiple types of data from the same cell, quantifies relationships between cell types, and identifies regulatory DNA elements responsible for gene activity.

CellWalker2 has been used to compare immune cell data from different studies and to map regulatory DNA regions in specific cell types. It can also compare brain cells across species, revealing both shared and unique cell types. The tool is open-source and available to researchers, providing a new way to interpret complex biological data and potentially linking disease risk variants to specific cell types and regulatory programs.