View Source

Differential expression analysis can be used to compare cell types. Here, we will compare malignant and oligodendrocyte cells to identify genes differentially regulated in malignant cells from the Oligodendroglioma subtype. This comparison is of interest because malignant cells in Oligodendroglioma are thought to originate from oligodendrocytes.

Normalize counts

To eliminate any zero values in the data set, add a small offset.

Select the Classified groups (green) data node
Select Normalization from the Normalization and scaling section of the task menu

The normalization options for single-cell RNA-Seq data are the same as for pooled or bulk RNA-Seq data.

Drag and drop Add from the left panel to the right panel
Set to 0.0001
Select Finish

Filter samples

To analyze only the Oligodendroglioma subtype, we can filter the samples.

Select the green Normalized counts data node
Select Filter samples from the Filtering section of the task menu
Set the filter to Include samples where Type is Oligodendroglioma
Select AND
Set the second filter to Exclude samples where Classification is Microglia
Select Finish to apply the filter

A Filtered Counts data node will be created with only cells that are from Oligodendroglioma samples and are classified as either malignant or oligodendrocyte.

Identify differentially expressed genes

Select the green Filtered counts data node
Select Detect differential expression (GSA)

The configuration options include cell-level attributes. Some of these are inherited from the sample level, like Type. Here, we want to compare only cells of a particular type from particular classification so we will include Type and Classification.

Select Type
Select Classification
Select Next

Next, we will set up a comparison between malignant cells and oligodendrocytes in Oligodendroglioma.

Select Malignant and Oligodendroglioma in the top panels
Select Oligodendrocytes in the bottom panels

Because we are analyzing sparse data, we need to relax the default low expression filter. By default the low expression filter is set a minimum of 1 average normalized read; here, we will turn it off.