TCellSelection¶

Separate T and non-T cells and select T cells.

If all of your cells are T cells, do not set any configurations for this process.

In such a case, SeuratClusteringOfAllCells should not be used, and SeuratClustering will be clustering all of the cells, which are all T cells.

There are two ways to separate T and non-T cells:

1. Use the an expression indicator directly from the metadata.
   
2. Use the expression values of indicator genes, and the clonotype percentage of the clusters.
   

You can also use indicator gene expression values only to select T cells by setting envs.ignore_tcr to true.

Environment Variables¶

• ignore_tcr (flag): Default: False.
  Ignore TCR information for T cell selection.
  Use only the expression values of indicator genes.
  In this case, the Clonotype_Pct column does not exist in the metadata.
  If you want to use k-means to select T cells, you must have more than 1 indicator gene, and the first indicator gene in envs.indicator_genes must be a positive marker, which will be used to select the cluster with higher expression values as T cells.
  
• tcell_selector: The expression passed to tidyseurat::mutate(is_TCell = ...) to indicate whether a cell is a T cell. For example, Clonotype_Pct > 0.25 to indicate cells with clonotype percentage > 25% are T cells.
  If indicator_genes is provided, the expression values can also be used in the expression. For example, Clonotype_Pct > 0.25 & CD3E > 0.
  If tcell_selector is not provided, a kmeans clustering will be performed on the expression values of indicator_genes and Clonotype_Pct, with K=2, and the cluster with higher clonotype percentage will be selected as T cells.
  

• indicator_genes (list): Default: ['CD3E'].
  A list of indicator genes whose expression values and clonotype percentage will be used to determine T cells.
  The markers could be either positive, such as CD3E, CD3D, CD3G, or negative, such as CD19, CD14, CD68.
  

• kmeans (type=json): Default: {'nstart': 25}.
  The parameters for kmeans clustering.
  Other arguments for stats::kmeans can be provided here. If there are dots in the argument names, replace them with -.
  

Examples¶

Use T cell indicator directly¶

If you have a metadata like this:

With the configuration below:

[TCellSelection.envs]
tcell_selector = "Clonotype_Pct > 0.25"

The T cells will be selected as:

Use indicator genes¶

Let's say we set the indicator genes to ["CD3D", "CD3E", "CD3G"].

The mean expression values will be calculated for each cluster:

Then a kmeans clustering will be performed on the mean expression values of the indicator genes, together with Clonotype_Pct, with K=2.

The cluster with higher clonoype percentage will be selected as T cells (is_TCell = TRUE), and sent to SeuratClustering for further clustering and downstream analysis.