Question: How do I create cell clusters using the Feature Plot in Loupe Browser?
Answer: The Feature Plot view enables to plot the expression level of one or two features for each cell barcode, which makes it easy to threshold cell population by expression levels, in particular for features with high dynamic ranges, such as cell surface markers.
Here we are going to use the PBMC Gene Expression + Antibody Barcode .cloupe file to demonstrate the functionalities of the Feature Plot. This dataset was generated using approximately 8,000 PBMCs labeled with a panel of 17 TotalSeq™-C antibodies.
We switch the view from “Categories” to “Gene/Feature Expression” and t-SNE projection to Feature Plot:
To identify CD3+ T cells using the CD3_TotalSeq-C surface marker expression level, we enter CD3_TotalSeq-C in the Y-axis search bar. Next, we change the scale of the Y-axis to log (n+1). Representative graph below:
There are several inflection points in this curve. Around the 4200th cell on the X-axis, the UMI counts increases from approximately 100 to more than 500, which may be used as a representative threshold for CD3+ T cells in regard to the expression of this surface marker.
A second feature can now be plotted on the X-axis: for example, we add CD4_TotalSeq-C on the X axis and change the scale to log (n+1). In addition, we can add a third feature in the Active Feature List, for instance CD8a_TotalSeq-C, and identify cells positive for this antibody:
Cells in the top right corner of this plot would be positive for both CD3 and CD4; cells in the bottom right corner would be positive for CD4 only, whereas cells in the top left corner would express both CD3 and CD8a. Finally, cells at the bottom left corner would not express any of these markers.
Isotype controls are not expected to label any particular cell type as they lack specificity for epitopes. Therefore, isotype controls are a type of negative control that can be used to determine the level of nonspecific background caused by primary antibodies.
For example, the signal intensity of the isotype control IgG1_control_TotalSeqC appears to be fairly homogeneous across all cells in the sample.
The heatmap scale for IgG1_control_TotalSeqC ranges from 0.0 to 11.0. We can use the signal of this isotype control to set a threshold for TotalSeqC antibodies. For example, we can increase the minimum value of the heatmap scale to about half of the maximum value (6.0):
The vast majority of the cells no longer show the signal of the isotype control. It is now possible to better identify CD8_TotalseqC positive cells in the Feature Plot view:
This assignment mechanism allows for more accurate filtering than using the t-SNE plots. Different cell populations can be easily selected using the lasso tool to create two (or more) new categories: for this exercise we are going to create one category for CD3/CD4 positive cells, and second one for CD3/CD8 positive cells.
By switching the view to the Antibody t-SNE projection we can clearly see where these two populations exist in the t-SNE view:
Integration with Immune Profiling
It is possible to explore clonotypes in the newly generated clusters by importing the .vloupe file associated with this dataset:
Next, we select Custer in the Filter selector and search for CD3/CD4. Finally we choose the option “Select All Clonotypes” from the All Clonotype stack menu. CD3 and CD4 positive cells would be highlighted in dark blue in the t-SNE view: