2023.
Press F11 for full screen mode

CyTOF FCS files

  • Must be normalized and debarcoded.
  • All fcs files must have the same channels. It’s fixable though, when missing channels are not important ones.
  • Pre-gating by using a few markers is recommended.

Batch Correction - CyCombine

Pre batch correction

Fig. PCA - PRE

Fig. PCA - PRE

Shows batch-wise localization

Post batch correction

Fig. PCA - POST

Fig. PCA - POST

localization improved

Prepare panel information and meta data

metadata

panel list

A metadata and a panel list can be read from a csv(xslx) file or can be constructed as a data frame.

Read FCSs into SCE - CATALYST

sce <- prepData(data_folder, panel = panel_list, md = metadata, 
features = panel_list$fcs_colname)

CATALYST library comes with useful internal commands for data manupulation/extraction and plot production.

Data exploration


Check marker expression profiles to see whether markers were active.

Data exploration - scatter plots

Cofactor selection

Clustering depending on cofactor selection

UMAP by uwot to extract fgraph followed by leiden clustering

Umap

uwot::umap(matrix_4_umap, n_neighbors = 15, n_components = comp_numb, 
           min_dist = mdist, n_threads = 32, n_epochs = 200, ret_nn = TRUE,
           ret_extra = "fgraph")

Leiden

leiden(uwot_graph, partition_type = "ModularityVertexPartition", 
       weights = E(uwot_graph)$weight, resolution_parameter = rp)

UMAP color-coded by Leiden clustering or by markers

R vs Python

MST with type markers

MST with state markers

Heatmap for subtype annotation

Marker expressions in subtypes

cluster- sample- wise data.

Compare cluster-wise data in groups separated by metadata.

Feed extracted data for further analysis and modeling I

Feed extracted data for further analysis and modeling II