This vignette shows how to apply CellChat to the comparison analysis of multiple datasets with different cell type compositions. Almost all the CellChat’s functionalities can be applied.

Load the required libraries

library(CellChat)
library(ggplot2)                  
library(patchwork)
library(igraph)

Part I: Comparison analysis of multiple datasets with slightly different cell type compositions

For the datasets with slightly different cell type (group) compositions, CellChat can lift up the cell groups to the same cell labels across all datasets using the function liftCellChat, and then perform comparison analysis as the joint analysis of datasets with the same cell type compositions.

Here we take an example of comparison analysis of two embryonic mouse skin scRNA-seq datasets from days E13.5 and E14.5. There are 11 shared skin cell populations at E13.5 and E14.5 and additional two populations (i.e., dermal DC and pericytes) specific to E14.5. Therefore, we will lift up the cell groups from E13.5 to the same cell labels as E14.5.

Load CellChat object of each dataset

USERS need to run CellChat on each dataset seperately and then merge different CellChat objects together. Here we also do updateCellChat because these two objects are obtained using the earlier version (< 1.6.0) of CellChat.

#cellchat.E13 <- readRDS(url("https://ndownloader.figshare.com/files/25957094"))
cellchat.E13 <- readRDS("/Users/jinsuoqin/Documents/CellChat/tutorial/cellchat_embryonic_E13.rds")
cellchat.E13 <- updateCellChat(cellchat.E13)
#> Update slot 'var.features' from a vector to a list
#cellchat.E14 <- readRDS(url("https://ndownloader.figshare.com/files/25957634"))
cellchat.E14 <- readRDS("/Users/jinsuoqin/Documents/CellChat/tutorial/cellchat_embryonic_E14.rds")
cellchat.E14 <- updateCellChat(cellchat.E14)
#> Update slot 'var.features' from a vector to a list

Lift up CellChat object and merge together

Since there are additional two populations (i.e., dermal DC and pericytes) specific to E14.5 compared to E13.5, we lift up cellchat.E13 by lifting up the cell groups to the same cell labels as E14.5. liftCellChat will only update the slot related to the cell-cell communication network, including slots , and .

# Define the cell labels to lift up
group.new = levels(cellchat.E14@idents)
cellchat.E13 <- liftCellChat(cellchat.E13, group.new)
#> The CellChat object will be lifted up using the cell labels FIB-A, FIB-B, FIB-P, DC, Pericyte, MYL, Immune, ENDO, Muscle, MELA, Basal-P, Basal, Spinious
#> Update slots object@net, object@netP, object@idents in a single dataset...
object.list <- list(E13 = cellchat.E13, E14 = cellchat.E14)
cellchat <- mergeCellChat(object.list, add.names = names(object.list), cell.prefix = TRUE)
#> Warning in mergeCellChat(object.list, add.names = names(object.list),
#> cell.prefix = TRUE): Prefix cell names!
#> The cell barcodes in merged 'meta' is  rep1_AAACCTGCACCAACCG rep1_AAACGGGAGCCGATTT rep1_AAACGGGAGTATCGAA rep1_AAACGGGCATCTCCCA rep1_AAAGATGCACTTGGAT rep1_AAAGATGCAGTTCATG
#> Warning in mergeCellChat(object.list, add.names = names(object.list), cell.prefix = TRUE): The cell barcodes in merged 'meta' is different from those in the used data matrix.
#>               We now simply assign the colnames in the data matrix to the rownames of merged 'mata'!
#> Merge the following slots: 'data.signaling','images','net', 'netP','meta', 'idents', 'var.features' , 'DB', and 'LR'.

Visualize the inferred signaling network using the lifted object

# Hierarchy plot
pathways.show <- c("WNT") 
weight.max <- getMaxWeight(object.list, slot.name = c("netP"), attribute = pathways.show) # control the edge weights across different datasets
vertex.receiver = seq(1,10) # Left portion of hierarchy plot the shows signaling to dermal cells and right portion shows signaling to epidermal cells
par(mfrow = c(1,2), xpd=TRUE)
for (i in 1:length(object.list)) {
  netVisual_aggregate(object.list[[i]], signaling = pathways.show, vertex.receiver = vertex.receiver, edge.weight.max = weight.max[1], edge.width.max = 10, signaling.name = paste(pathways.show, names(object.list)[i]))
}

# Circle plot
pathways.show <- c("WNT") 
weight.max <- getMaxWeight(object.list, slot.name = c("netP"), attribute = pathways.show) # control the edge weights across different datasets
par(mfrow = c(1,2), xpd=TRUE)
for (i in 1:length(object.list)) {
  netVisual_aggregate(object.list[[i]], signaling = pathways.show, layout = "circle", edge.weight.max = weight.max[1], edge.width.max = 10, signaling.name = paste(pathways.show, names(object.list)[i]))
}

# Chord diagram
pathways.show <- c("WNT") 
par(mfrow = c(1,2), xpd=TRUE)
for (i in 1:length(object.list)) {
  netVisual_aggregate(object.list[[i]], signaling = pathways.show, layout = "chord", signaling.name = paste(pathways.show, names(object.list)[i]))
}

For the chord diagram, CellChat has an independent function netVisual_chord_cell to flexibly visualize the signaling network by adjusting different parameters in the circlize package. For example, we can define a named char vector group to create multiple-group chord diagram, e.g., grouping cell clusters into different cell types.

# Chord diagram
group.merged <- c(rep("Dermal", 10), rep("Epidermal", 3)) # grouping cell clusters into dermal and epidermal cells to study the cell-cell communication between dermal and epidermal
names(group.merged) <- levels(object.list[[1]]@idents)
pathways.show <- c("WNT") 
par(mfrow = c(1,2), xpd=TRUE)
for (i in 1:length(object.list)) {
  netVisual_chord_cell(object.list[[i]], signaling = pathways.show, group = group.merged, title.name = paste0(pathways.show, " signaling network - ", names(object.list)[i]))
}
#> Plot the aggregated cell-cell communication network at the signaling pathway level
#> Plot the aggregated cell-cell communication network at the signaling pathway level

Part II: Comparison analysis of multiple datasets with vastly different cell type compositions

CellChat can be used to compare cell-cell communication patterns between two scRNA-seq datasets from vastly distinct biological contexts – embryonic morphogenesis vs. wound-induced repair. For the datasets with vastly different cell type (group) compositions, most of the CellChat’s functionalities can be applied, except for the following two aspects:

However, USERS can still use computeNetSimilarityPairwise(cellchat, type = "structural") to perform structure similarity analysis.

Please check out our paper for the comparison analysis of two scRNA-seq datasets, one from embryonic day E13.5 skin and another from adult day 12 wound skin.

sessionInfo()
#> R version 4.1.2 (2021-11-01)
#> Platform: x86_64-apple-darwin17.0 (64-bit)
#> Running under: macOS Big Sur 10.16
#> 
#> Matrix products: default
#> BLAS:   /Library/Frameworks/R.framework/Versions/4.1/Resources/lib/libRblas.0.dylib
#> LAPACK: /Library/Frameworks/R.framework/Versions/4.1/Resources/lib/libRlapack.dylib
#> 
#> locale:
#> [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
#> 
#> attached base packages:
#> [1] stats     graphics  grDevices utils     datasets  methods   base     
#> 
#> other attached packages:
#> [1] patchwork_1.1.1     CellChat_1.6.0      Biobase_2.54.0     
#> [4] BiocGenerics_0.40.0 ggplot2_3.3.5       igraph_1.3.4       
#> [7] dplyr_1.0.7        
#> 
#> loaded via a namespace (and not attached):
#>   [1] colorspace_2.0-2      ggsignif_0.6.3        rjson_0.2.20         
#>   [4] ellipsis_0.3.2        circlize_0.4.13       GlobalOptions_0.1.2  
#>   [7] BiocNeighbors_1.12.0  clue_0.3-60           ggpubr_0.4.0         
#>  [10] listenv_0.8.0         ggrepel_0.9.1         RSpectra_0.16-0      
#>  [13] fansi_0.5.0           codetools_0.2-18      doParallel_1.0.16    
#>  [16] knitr_1.40            jsonlite_1.7.2        broom_0.7.10         
#>  [19] gridBase_0.4-7        cluster_2.1.2         png_0.1-7            
#>  [22] compiler_4.1.2        backports_1.4.1       assertthat_0.2.1     
#>  [25] Matrix_1.3-4          fastmap_1.1.0         htmltools_0.5.2      
#>  [28] tools_4.1.2           coda_0.19-4           gtable_0.3.0         
#>  [31] glue_1.6.0            reshape2_1.4.4        Rcpp_1.0.7           
#>  [34] carData_3.0-4         statnet.common_4.5.0  jquerylib_0.1.4      
#>  [37] NMF_0.23.0            vctrs_0.3.8           svglite_2.0.0        
#>  [40] iterators_1.0.13      ggalluvial_0.12.3     xfun_0.33            
#>  [43] stringr_1.4.0         globals_0.14.0        network_1.17.1       
#>  [46] lifecycle_1.0.1       irlba_2.3.5           rngtools_1.5.2       
#>  [49] rstatix_0.7.0         future_1.23.0         scales_1.1.1         
#>  [52] parallel_4.1.2        RColorBrewer_1.1-2    ComplexHeatmap_2.10.0
#>  [55] yaml_2.2.1            reticulate_1.22       pbapply_1.5-0        
#>  [58] pkgmaker_0.32.2       ggnetwork_0.5.10      sass_0.4.0           
#>  [61] stringi_1.7.6         highr_0.9             S4Vectors_0.32.3     
#>  [64] foreach_1.5.1         BiocParallel_1.28.3   shape_1.4.6          
#>  [67] rlang_0.4.12          pkgconfig_2.0.3       systemfonts_1.0.2    
#>  [70] matrixStats_0.61.0    evaluate_0.17         lattice_0.20-45      
#>  [73] purrr_0.3.4           cowplot_1.1.1         tidyselect_1.1.1     
#>  [76] parallelly_1.30.0     plyr_1.8.6            magrittr_2.0.1       
#>  [79] R6_2.5.1              IRanges_2.28.0        generics_0.1.1       
#>  [82] sna_2.6               DBI_1.1.2             pillar_1.6.4         
#>  [85] withr_2.4.3           abind_1.4-5           tibble_3.1.6         
#>  [88] future.apply_1.8.1    crayon_1.4.2          car_3.0-12           
#>  [91] utf8_1.2.2            rmarkdown_2.17        GetoptLong_1.0.5     
#>  [94] grid_4.1.2            FNN_1.1.3             digest_0.6.29        
#>  [97] xtable_1.8-4          tidyr_1.1.4           stats4_4.1.2         
#> [100] munsell_0.5.0         registry_0.5-1        bslib_0.3.1