Visualization and Quantification of High-Dimensional Cytometry Data using Cytofast and the Upstream Clustering Methods FlowSOM and Cytosplore

The data generated by mass cytometry are complex and need to be visualized in an efficient way and simple manner. Cytofast is a method which highlights immunological patterns within the immune cell population and determines cell subsets which are linked to clinical treatments or experimental groups. Cytofast can be applied after any crystalling method like FlowSOM or Cytosplore.

It will allow you to discover a cell subset which are linked to clinical treatment or an experimental group. With this method, you will be able to see the immunological overview of the data at a glance in a quantitative manner. Begin by creating clusters with either Cytosplore or FlowSOM.

If using Cytosplore, upload the FCS files by clicking Files and open FCS files. Then select a cofactor for Hyperbolic ArcSinh transformation when prompted and click on add unique sample tag as channel. Select run HSNE to run an HSNE level of three and wait for the map to be generated.

On the first HSNE level, check the cells that are positive for CD161 by selecting the CD161 positive cells and right-clicking zoom into selection. At the second level, repeat the procedure to reach the third level with only CD161 positive events. Once the last TSNE map is generated, save the clusters defined by Cytosplore by right-clicking on the map and choosing save clusters.

Choose the directory of the output files as prompted by Cytosplore and note this location because it will later be used to load FSC files into R.Use a simple characters only name when renaming the output files which will make identification and further handling easier and select save. Load the files into R with the designated function readcytosploreFCS. To clean the data, remove some parameters such as time and background by checking the position of the column related to its unnecessary parameters and removing it from the matrix.

Next, reorder the markers so that lineage markers are displayed first followed by functional markers. Link the metadata file to the generated data from Cytosplore by uploading the spreadsheet metafile containing clinical information. To perform clustering by FlowSOM, load the raw data that was previously gated on CD161 positive events in R with the read.

flowSet function. Select the relevant biological markers by selecting the proper columns and transforming the data in an ArcSinh5 manner. Apply a cofactor of five by choosing cofacter=5 in the function.

Cluster the data using the FlowSOM function and compare FlwoSOM and Cytosplore by choosing to cluster the data in 10 subsets same as the output previously yielded by Cytosplore. Then assign each cell to its identified subset and sample ID.Load the metadata file in R containing the group assignment and link it to the FCS files using the code from the text manuscript. Create a CF list based on the dataframe obtained from FlowSOM.

Then reorder the markers to appear similarly to the output from the Cytosplore analysis. Prior to creating the heat maps, generate the count table per sample by using the function cellCount. Since some clusters contain fewer cells than others, scale the data per cluster by specifying scale=true inside the function which makes it easy to see the dispersion among samples.

Visualize the data with a heat map then visualize with box plots by generating the cell count but not scaling the data to obtain the frequency of each cluster. Finally, visualize the expression intensity of the CD45, CD11c and CD54 markers by the noting marker names and including them in the MSI plot function. Cytofast runs several possible outputs including the heat map of all clusters identified in the analysis and based on marker expression.

The dendrogram on the top represents the hierarchical similarity between the identified clusters. The upper panel displays another heat map showing the relative quantity of corresponding subsets in each sample. Meanwhile, the dendrogram on the right shows the similarity between samples based on subset frequencies demonstrating that the phenotype of natural killer cells is shaped by PD-L1 treatment three days after injection.

The combined heat maps can be obtained for FlowSOM followed by Cytofast and for Cytosplore followed by Cytofast. Cytofast can also be used to present the data quantitatively and display the results in box plots. Another feature that is included in the Cytofast package is the MSI plot function that shows the median signal intensity plot of every marker per group.

This function allows detection of global changes such as increases in the expression of CD54 or CD11c in natural killer cells of the PD-L1 treated group. This technique is a rapid way to visualize and quantify data and can be used to discover novel cellular responses after therapy. After this method, the global test package in R can be used to test a group of covariates for association with response variables.

This will show the correlation between each subsets and the clinical outcome.