Single-cell genomics gives researchers the power to interrogate the cell state of individual cells within a sample. This allows us to ask questions like, “how are different cell types affected by HIV?” and “which genetic programs are activated in cancer cells?” In investigating these questions, researchers often seek out differences in genetic expression across different groups of cells. To make it easier to spot these differences, Single Cell Portal now supports exploratory differential expression (DE) analyses for some datasets. This tool will help our users get a quick impression of the genes that distinguish between cell groups of interest, which can help verify cell annotations and inspire more rigorous follow-up work.

How does it work?

When you visit a study where this tool is available, you’ll see a “Differential Expression” button on the right-hand panel of the study’s ‘Explore’ tab.

After clicking that button, you can select a target annotation label (group of cells) that you want to compare to the rest of the cells in the data. 

The options for the target group are based on the “annotation” selected from the drop-down menu just above the “Differential Expression” button on the previous screen. If you want to compare across a different kind of annotation – for example, if you want to compare across disease states rather than cell types – you can return to that previous screen and select a different annotation.

Once you’ve selected an annotation label (in this example, B cells), you’ll see a table of results: the 15 genes that are expressed most differently in B cells versus all other cell types. The genes are listed in order of statistical significance. Each row lists its gene’s log2 fold change (a measure of how much more the gene was expressed in the target cells than the rest of the cells) and the corresponding FDR-corrected p-value. You can see more details on how we’re generating these results in SCP's differential expression documentation.

To understand how a gene’s expression varies across the cells in the data, you can click on the empty circle next to its row (box #1 in the image below). This will generate a gene expression scatterplot (box #2): a plot of the clustered data in a low-dimensional embedding, where each point corresponds to one cell and the cells are colored according to the selected gene’s measured expression.