This vignette outlines the steps of inference, analysis and visualization of cell-cell communication network for a single dataset using CellChat. We showcase CellChat’s diverse functionalities by applying it to a scRNA-seq data on cells from lesional (LS, diseased) human skin from patients.

CellChat requires gene expression data of cells as the user input and models the probability of cell-cell communication by integrating gene expression with prior knowledge of the interactions between signaling ligands, receptors and their cofactors.

Upon infering the intercellular communication network, CellChat provides functionality for further data exploration, analysis, and visualization.

Load the required libraries

library(CellChat)
library(patchwork)
options(stringsAsFactors = FALSE)

Part I: Data input & processing and initialization of CellChat object

CellChat requires two user inputs: one is the gene expression data of cells, and the other is either user assigned cell labels (i.e., label-based mode) or a low-dimensional representation of the single-cell data (i.e., label-free mode). For the latter, CellChat automatically groups cells by building a shared neighbor graph based on the cell-cell distance in the low-dimensional space or the pseudotemporal trajectory space.

Load data

For the gene expression data matrix, genes should be in rows with rownames and cells in columns with colnames. Normalized data (e.g., library-size normalization and then log-transformed with a pseudocount of 1) is required as input for CellChat analysis. If user provides count data, we provide a normalizeData function to account for library size and then do log-transformed. For the cell group information, a dataframe with rownames is required as input for CellChat.

# Here we load a scRNA-seq data matrix and its associated cell meta data
#load(url("https://ndownloader.figshare.com/files/25950872")) # This is a combined data from two biological conditions: normal and diseases
load("/Users/jinsuoqin/Documents/CellChat/tutorial/data_humanSkin_CellChat.rda")
data.input = data_humanSkin$data # normalized data matrix
meta = data_humanSkin$meta # a dataframe with rownames containing cell mata data
cell.use = rownames(meta)[meta$condition == "LS"] # extract the cell names from disease data

# Prepare input data for CelChat analysis
data.input = data.input[, cell.use]
meta = meta[cell.use, ]
# meta = data.frame(labels = meta$labels[cell.use], row.names = colnames(data.input)) # manually create a dataframe consisting of the cell labels
unique(meta$labels) # check the cell labels
#>  [1] Inflam. FIB  FBN1+ FIB    APOE+ FIB    COL11A1+ FIB cDC2        
#>  [6] LC           Inflam. DC   cDC1         CD40LG+ TC   Inflam. TC  
#> [11] TC           NKT         
#> 12 Levels: APOE+ FIB FBN1+ FIB COL11A1+ FIB Inflam. FIB cDC1 cDC2 ... NKT

Create a CellChat object

USERS can create a new CellChat object from a data matrix, Seurat or SingleCellExperiment object. If input is a Seurat or SingleCellExperiment object, the meta data in the object will be used by default and USER must provide group.by to define the cell groups. e.g, group.by = “ident” for the default cell identities in Seurat object.

NB: If USERS load previously calculated CellChat object (version < 0.5.0), please update the object via updateCellChat

cellchat <- createCellChat(object = data.input, meta = meta, group.by = "labels")
#> [1] "Create a CellChat object from a data matrix"
#> Set cell identities for the new CellChat object 
#> The cell groups used for CellChat analysis are  APOE+ FIB FBN1+ FIB COL11A1+ FIB Inflam. FIB cDC1 cDC2 LC Inflam. DC TC Inflam. TC CD40LG+ TC NKT

Add cell information into meta slot of the object (Optional)

If cell mata information is not added when creating CellChat object, USERS can also add it later using addMeta, and set the default cell identities using setIdent.

cellchat <- addMeta(cellchat, meta = meta)
cellchat <- setIdent(cellchat, ident.use = "labels") # set "labels" as default cell identity
levels(cellchat@idents) # show factor levels of the cell labels
groupSize <- as.numeric(table(cellchat@idents)) # number of cells in each cell group

Set the ligand-receptor interaction database

Our database CellChatDB is a manually curated database of literature-supported ligand-receptor interactions in both human and mouse. CellChatDB in mouse contains 2,021 validated molecular interactions, including 60% of secrete autocrine/paracrine signaling interactions, 21% of extracellular matrix (ECM)-receptor interactions and 19% of cell-cell contact interactions. CellChatDB in human contains 1,939 validated molecular interactions, including 61.8% of paracrine/autocrine signaling interactions, 21.7% of extracellular matrix (ECM)-receptor interactions and 16.5% of cell-cell contact interactions.

Users can update CellChatDB by adding their own curated ligand-receptor pairs.Please check our tutorial on how to do it.

CellChatDB <- CellChatDB.human # use CellChatDB.mouse if running on mouse data
showDatabaseCategory(CellChatDB)

# Show the structure of the database
dplyr::glimpse(CellChatDB$interaction)
#> Rows: 1,939
#> Columns: 11
#> $ interaction_name   <chr> "TGFB1_TGFBR1_TGFBR2", "TGFB2_TGFBR1_TGFBR2", "TGFB…
#> $ pathway_name       <chr> "TGFb", "TGFb", "TGFb", "TGFb", "TGFb", "TGFb", "TG…
#> $ ligand             <chr> "TGFB1", "TGFB2", "TGFB3", "TGFB1", "TGFB1", "TGFB2…
#> $ receptor           <chr> "TGFbR1_R2", "TGFbR1_R2", "TGFbR1_R2", "ACVR1B_TGFb…
#> $ agonist            <chr> "TGFb agonist", "TGFb agonist", "TGFb agonist", "TG…
#> $ antagonist         <chr> "TGFb antagonist", "TGFb antagonist", "TGFb antagon…
#> $ co_A_receptor      <chr> "", "", "", "", "", "", "", "", "", "", "", "", "",…
#> $ co_I_receptor      <chr> "TGFb inhibition receptor", "TGFb inhibition recept…
#> $ evidence           <chr> "KEGG: hsa04350", "KEGG: hsa04350", "KEGG: hsa04350…
#> $ annotation         <chr> "Secreted Signaling", "Secreted Signaling", "Secret…
#> $ interaction_name_2 <chr> "TGFB1 - (TGFBR1+TGFBR2)", "TGFB2 - (TGFBR1+TGFBR2)…

# use a subset of CellChatDB for cell-cell communication analysis
CellChatDB.use <- subsetDB(CellChatDB, search = "Secreted Signaling") # use Secreted Signaling
# use all CellChatDB for cell-cell communication analysis
# CellChatDB.use <- CellChatDB # simply use the default CellChatDB

# set the used database in the object
cellchat@DB <- CellChatDB.use

Preprocessing the expression data for cell-cell communication analysis

To infer the cell state-specific communications, we identify over-expressed ligands or receptors in one cell group and then identify over-expressed ligand-receptor interactions if either ligand or receptor is over-expressed.

We also provide a function to project gene expression data onto protein-protein interaction (PPI) network. Specifically, a diffusion process is used to smooth genes’ expression values based on their neighbors’ defined in a high-confidence experimentally validated protein-protein network. This function is useful when analyzing single-cell data with shallow sequencing depth because the projection reduces the dropout effects of signaling genes, in particular for possible zero expression of subunits of ligands/receptors. One might be concerned about the possible artifact introduced by this diffusion process, however, it will only introduce very weak communications. USERS can also skip this step and set raw.use = TRUE in the function computeCommunProb().

# subset the expression data of signaling genes for saving computation cost
cellchat <- subsetData(cellchat) # This step is necessary even if using the whole database
future::plan("multiprocess", workers = 4) # do parallel
#> Warning: Strategy 'multiprocess' is deprecated in future (>= 1.20.0). Instead,
#> explicitly specify either 'multisession' or 'multicore'. In the current R
#> session, 'multiprocess' equals 'multisession'.
#> Warning in supportsMulticoreAndRStudio(...): [ONE-TIME WARNING] Forked
#> processing ('multicore') is not supported when running R from RStudio
#> because it is considered unstable. For more details, how to control forked
#> processing or not, and how to silence this warning in future R sessions, see ?
#> parallelly::supportsMulticore
cellchat <- identifyOverExpressedGenes(cellchat)
cellchat <- identifyOverExpressedInteractions(cellchat)

# project gene expression data onto PPI (Optional: when running it, USER should set `raw.use = FALSE` in the function `computeCommunProb()` in order to use the projected data)
# cellchat <- projectData(cellchat, PPI.human)

Part II: Inference of cell-cell communication network

CellChat infers the biologically significant cell-cell communication by assigning each interaction with a probability value and peforming a permutation test. CellChat models the probability of cell-cell communication by integrating gene expression with prior known knowledge of the interactions between signaling ligands, receptors and their cofactors using the law of mass action.

The number of inferred ligand-receptor pairs clearly depends on the method for calculating the average gene expression per cell group. By default, CellChat uses a statistically robust mean method called ‘trimean’, which produces fewer interactions than other methods. However, we find that CellChat performs well at predicting stronger interactions, which is very helpful for narrowing down on interactions for further experimental validations. In computeCommunProb, we provide an option for using other methods, such as 5% and 10% truncated mean, to calculating the average gene expression. Of note, ‘trimean’ approximates 25% truncated mean, implying that the average gene expression is zero if the percent of expressed cells in one group is less than 25%. To use 10% truncated mean, USER can set type = "truncatedMean" and trim = 0.1. The function computeAveExpr can help to check the average expression of signaling genes of interest, e.g, computeAveExpr(cellchat, features = c("CXCL12","CXCR4"), type = "truncatedMean", trim = 0.1).

When analyzing unsorted single-cell transcriptomes, under the assumption that abundant cell populations tend to send collectively stronger signals than the rare cell populations, CellChat can also consider the effect of cell proportion in each cell group in the probability calculation. USER can set population.size = TRUE.

Compute the communication probability and infer cellular communication network

If well-known signaling pathways in the studied biological process are not predicted, USER can try truncatedMean to change the method for calculating the average gene expression per cell group.

cellchat <- computeCommunProb(cellchat)
#> triMean is used for calculating the average gene expression per cell group. 
#> [1] ">>> Run CellChat on sc/snRNA-seq data <<< [2022-11-26 08:34:38]"
#> [1] ">>> CellChat inference is done. Parameter values are stored in `object@options$parameter` <<< [2022-11-26 08:35:36]"
# Filter out the cell-cell communication if there are only few number of cells in certain cell groups
cellchat <- filterCommunication(cellchat, min.cells = 10)

Extract the inferred cellular communication network as a data frame

We provide a function subsetCommunication to easily access the inferred cell-cell communications of interest. For example,

  • df.net <- subsetCommunication(cellchat) returns a data frame consisting of all the inferred cell-cell communications at the level of ligands/receptors. Set slot.name = "netP" to access the the inferred communications at the level of signaling pathways

  • df.net <- subsetCommunication(cellchat, sources.use = c(1,2), targets.use = c(4,5)) gives the inferred cell-cell communications sending from cell groups 1 and 2 to cell groups 4 and 5.

  • df.net <- subsetCommunication(cellchat, signaling = c("WNT", "TGFb")) gives the inferred cell-cell communications mediated by signaling WNT and TGFb.

Infer the cell-cell communication at a signaling pathway level

CellChat computes the communication probability on signaling pathway level by summarizing the communication probabilities of all ligands-receptors interactions associated with each signaling pathway.

NB: The inferred intercellular communication network of each ligand-receptor pair and each signaling pathway is stored in the slot ‘net’ and ‘netP’, respectively.

cellchat <- computeCommunProbPathway(cellchat)

Calculate the aggregated cell-cell communication network

We can calculate the aggregated cell-cell communication network by counting the number of links or summarizing the communication probability. USER can also calculate the aggregated network among a subset of cell groups by setting sources.use and targets.use.

cellchat <- aggregateNet(cellchat)

We can also visualize the aggregated cell-cell communication network. For example, showing the number of interactions or the total interaction strength (weights) between any two cell groups using circle plot.

groupSize <- as.numeric(table(cellchat@idents))
par(mfrow = c(1,2), xpd=TRUE)
netVisual_circle(cellchat@net$count, vertex.weight = groupSize, weight.scale = T, label.edge= F, title.name = "Number of interactions")
netVisual_circle(cellchat@net$weight, vertex.weight = groupSize, weight.scale = T, label.edge= F, title.name = "Interaction weights/strength")

Due to the complicated cell-cell communication network, we can examine the signaling sent from each cell group. Here we also control the parameter edge.weight.max so that we can compare edge weights between differet networks.

mat <- cellchat@net$weight
par(mfrow = c(3,4), xpd=TRUE)
for (i in 1:nrow(mat)) {
  mat2 <- matrix(0, nrow = nrow(mat), ncol = ncol(mat), dimnames = dimnames(mat))
  mat2[i, ] <- mat[i, ]
  netVisual_circle(mat2, vertex.weight = groupSize, weight.scale = T, edge.weight.max = max(mat), title.name = rownames(mat)[i])
}

Part III: Visualization of cell-cell communication network

Upon infering the cell-cell communication network, CellChat provides various functionality for further data exploration, analysis, and visualization.

Visualize each signaling pathway using Hierarchy plot, Circle plot or Chord diagram

Hierarchy plot: USER should define vertex.receiver, which is a numeric vector giving the index of the cell groups as targets in the left part of hierarchy plot. This hierarchical plot consist of two components: the left portion shows autocrine and paracrine signaling to certain cell groups of interest (i.e, the defined vertex.receiver), and the right portion shows autocrine and paracrine signaling to the remaining cell groups in the dataset. Thus, hierarchy plot provides an informative and intuitive way to visualize autocrine and paracrine signaling communications between cell groups of interest. For example, when studying the cell-cell communication between fibroblasts and immune cells, USER can define vertex.receiver as all fibroblast cell groups.

Chord diagram: CellChat provides two functions netVisual_chord_cell and netVisual_chord_gene for visualizing cell-cell communication with different purposes and different levels. netVisual_chord_cell is used for visualizing the cell-cell communication between different cell groups (where each sector in the chord diagram is a cell group), and netVisual_chord_gene is used for visualizing the cell-cell communication mediated by mutiple ligand-receptors or signaling pathways (where each sector in the chord diagram is a ligand, receptor or signaling pathway.)

Explnations of edge color/weight, node color/size/shape: In all visualization plots, edge colors are consistent with the sources as sender, and edge weights are proportional to the interaction strength. Thicker edge line indicates a stronger signal. In the Hierarchy plot and Circle plot, circle sizes are proportional to the number of cells in each cell group. In the hierarchy plot, solid and open circles represent source and target, respectively. In the Chord diagram, the inner thinner bar colors represent the targets that receive signal from the corresponding outer bar. The inner bar size is proportional to the signal strength received by the targets. Such inner bar is helpful for interpreting the complex chord diagram. Note that there exist some inner bars without any chord for some cell groups, please just igore it because this is an issue that has not been addressed by circlize package.

Visualization of cell-cell communication at different levels: One can visualize the inferred communication network of signaling pathways using netVisual_aggregate, and visualize the inferred communication networks of individual L-R pairs associated with that signaling pathway using netVisual_individual.

Here we take input of one signaling pathway as an example. All the signaling pathways showing significant communications can be accessed by cellchat@netP$pathways.

pathways.show <- c("CXCL") 
# Hierarchy plot
# Here we define `vertex.receive` so that the left portion of the hierarchy plot shows signaling to fibroblast and the right portion shows signaling to immune cells 
vertex.receiver = seq(1,4) # a numeric vector. 
netVisual_aggregate(cellchat, signaling = pathways.show,  vertex.receiver = vertex.receiver)
# Circle plot
par(mfrow=c(1,1))
netVisual_aggregate(cellchat, signaling = pathways.show, layout = "circle")

# Chord diagram
par(mfrow=c(1,1))
netVisual_aggregate(cellchat, signaling = pathways.show, layout = "chord")

# Heatmap
par(mfrow=c(1,1))
netVisual_heatmap(cellchat, signaling = pathways.show, color.heatmap = "Reds")
#> Do heatmap based on a single object

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.cellType <- c(rep("FIB", 4), rep("DC", 4), rep("TC", 4)) # grouping cell clusters into fibroblast, DC and TC cells
names(group.cellType) <- levels(cellchat@idents)
netVisual_chord_cell(cellchat, signaling = pathways.show, group = group.cellType, title.name = paste0(pathways.show, " signaling network"))
#> Plot the aggregated cell-cell communication network at the signaling pathway level

Compute the contribution of each ligand-receptor pair to the overall signaling pathway and visualize cell-cell communication mediated by a single ligand-receptor pair

netAnalysis_contribution(cellchat, signaling = pathways.show)

We can also visualize the cell-cell communication mediated by a single ligand-receptor pair. We provide a function extractEnrichedLR to extract all the significant interactions (L-R pairs) and related signaling genes for a given signaling pathway.

pairLR.CXCL <- extractEnrichedLR(cellchat, signaling = pathways.show, geneLR.return = FALSE)
LR.show <- pairLR.CXCL[1,] # show one ligand-receptor pair
# Hierarchy plot
vertex.receiver = seq(1,4) # a numeric vector
netVisual_individual(cellchat, signaling = pathways.show,  pairLR.use = LR.show, vertex.receiver = vertex.receiver)
#> [[1]]
# Circle plot
netVisual_individual(cellchat, signaling = pathways.show, pairLR.use = LR.show, layout = "circle")

#> [[1]]
# Chord diagram
netVisual_individual(cellchat, signaling = pathways.show, pairLR.use = LR.show, layout = "chord")

#> [[1]]

Automatically save the plots of the all inferred network for quick exploration

In practical use, USERS can use ‘for … loop’ to automatically save the all inferred network for quick exploration using netVisual. netVisual supports an output in the formats of svg, png and pdf.

# Access all the signaling pathways showing significant communications
pathways.show.all <- cellchat@netP$pathways
# check the order of cell identity to set suitable vertex.receiver
levels(cellchat@idents)
vertex.receiver = seq(1,4)
for (i in 1:length(pathways.show.all)) {
  # Visualize communication network associated with both signaling pathway and individual L-R pairs
  netVisual(cellchat, signaling = pathways.show.all[i], vertex.receiver = vertex.receiver, layout = "hierarchy")
  # Compute and visualize the contribution of each ligand-receptor pair to the overall signaling pathway
  gg <- netAnalysis_contribution(cellchat, signaling = pathways.show.all[i])
  ggsave(filename=paste0(pathways.show.all[i], "_L-R_contribution.pdf"), plot=gg, width = 3, height = 2, units = 'in', dpi = 300)
}

Visualize cell-cell communication mediated by multiple ligand-receptors or signaling pathways

Bubble plot

We can also show all the significant interactions (L-R pairs) from some cell groups to other cell groups using netVisual_bubble.

# show all the significant interactions (L-R pairs) from some cell groups (defined by 'sources.use') to other cell groups (defined by 'targets.use')
netVisual_bubble(cellchat, sources.use = 4, targets.use = c(5:11), remove.isolate = FALSE)
#> Comparing communications on a single object

# show all the significant interactions (L-R pairs) associated with certain signaling pathways
netVisual_bubble(cellchat, sources.use = 4, targets.use = c(5:11), signaling = c("CCL","CXCL"), remove.isolate = FALSE)
#> Comparing communications on a single object

# show all the significant interactions (L-R pairs) based on user's input (defined by `pairLR.use`)
pairLR.use <- extractEnrichedLR(cellchat, signaling = c("CCL","CXCL","FGF"))
netVisual_bubble(cellchat, sources.use = c(3,4), targets.use = c(5:8), pairLR.use = pairLR.use, remove.isolate = TRUE)
#> Comparing communications on a single object

Chord diagram

Similar to Bubble plot, CellChat provides a function netVisual_chord_gene for drawing Chord diagram to

  • show all the interactions (L-R pairs or signaling pathways) from some cell groups to other cell groups. Two special cases: one is showing all the interactions sending from one cell groups and the other is showing all the interactions received by one cell group.

  • show the interactions inputted by USERS or certain signaling pathways defined by USERS

# show all the significant interactions (L-R pairs) from some cell groups (defined by 'sources.use') to other cell groups (defined by 'targets.use')
# show all the interactions sending from Inflam.FIB
netVisual_chord_gene(cellchat, sources.use = 4, targets.use = c(5:11), lab.cex = 0.5,legend.pos.y = 30)

# show all the interactions received by Inflam.DC
netVisual_chord_gene(cellchat, sources.use = c(1,2,3,4), targets.use = 8, legend.pos.x = 15)

# show all the significant interactions (L-R pairs) associated with certain signaling pathways
netVisual_chord_gene(cellchat, sources.use = c(1,2,3,4), targets.use = c(5:11), signaling = c("CCL","CXCL"),legend.pos.x = 8)

# show all the significant signaling pathways from some cell groups (defined by 'sources.use') to other cell groups (defined by 'targets.use')
netVisual_chord_gene(cellchat, sources.use = c(1,2,3,4), targets.use = c(5:11), slot.name = "netP", legend.pos.x = 10)

NB: Please ignore the note when generating the plot such as “Note: The first link end is drawn out of sector ‘MIF’.”. If the gene names are overlapped, you can adjust the argument small.gap by decreasing the value.

Plot the signaling gene expression distribution using violin/dot plot

We can plot the gene expression distribution of signaling genes related to L-R pairs or signaling pathway using a Seurat wrapper function plotGeneExpression.

plotGeneExpression(cellchat, signaling = "CXCL")
#> Registered S3 method overwritten by 'spatstat.geom':
#>   method     from
#>   print.boxx cli
#> Scale for 'y' is already present. Adding another scale for 'y', which will
#> replace the existing scale.
#> Scale for 'y' is already present. Adding another scale for 'y', which will
#> replace the existing scale.
#> Scale for 'y' is already present. Adding another scale for 'y', which will
#> replace the existing scale.

By default, plotGeneExpression only shows the expression of signaling genes related to the inferred significant communications. USERS can show the expression of all signaling genes related to one signaling pathway by

plotGeneExpression(cellchat, signaling = "CXCL", enriched.only = FALSE)

Alternatively, USERS can extract the signaling genes related to the inferred L-R pairs or signaling pathway using extractEnrichedLR, and then plot gene expression using Seurat package.

Part IV: Systems analysis of cell-cell communication network

To facilitate the interpretation of the complex intercellular communication networks, CellChat quantitively measures networks through methods abstracted from graph theory, pattern recognition and manifold learning.

Identify signaling roles (e.g., dominant senders, receivers) of cell groups as well as the major contributing signaling

CellChat allows ready identification of dominant senders, receivers, mediators and influencers in the intercellular communication network by computing several network centrality measures for each cell group. Specifically, we used measures in weighted-directed networks, including out-degree, in-degree, flow betweenesss and information centrality, to respectively identify dominant senders, receivers, mediators and influencers for the intercellular communications. In a weighteddirected network with the weights as the computed communication probabilities, the outdegree, computed as the sum of communication probabilities of the outgoing signaling from a cell group, and the in-degree, computed as the sum of the communication probabilities of the incoming signaling to a cell group, can be used to identify the dominant cell senders and receivers of signaling networks, respectively. For the definition of flow betweenness and information centrality, please check our paper and related reference.

Compute and visualize the network centrality scores

# Compute the network centrality scores
cellchat <- netAnalysis_computeCentrality(cellchat, slot.name = "netP") # the slot 'netP' means the inferred intercellular communication network of signaling pathways
# Visualize the computed centrality scores using heatmap, allowing ready identification of major signaling roles of cell groups
netAnalysis_signalingRole_network(cellchat, signaling = pathways.show, width = 8, height = 2.5, font.size = 10)

Visualize the dominant senders (sources) and receivers (targets) in a 2D space

We also provide another intutive way to visualize the dominant senders (sources) and receivers (targets) in a 2D space using scatter plot.

# Signaling role analysis on the aggregated cell-cell communication network from all signaling pathways
gg1 <- netAnalysis_signalingRole_scatter(cellchat)
#> Signaling role analysis on the aggregated cell-cell communication network from all signaling pathways
# Signaling role analysis on the cell-cell communication networks of interest
gg2 <- netAnalysis_signalingRole_scatter(cellchat, signaling = c("CXCL", "CCL"))
#> Signaling role analysis on the cell-cell communication network from user's input
gg1 + gg2

Identify signals contributing most to outgoing or incoming signaling of certain cell groups

We can also answer the question on which signals contributing most to outgoing or incoming signaling of certain cell groups.

# Signaling role analysis on the aggregated cell-cell communication network from all signaling pathways
ht1 <- netAnalysis_signalingRole_heatmap(cellchat, pattern = "outgoing")
ht2 <- netAnalysis_signalingRole_heatmap(cellchat, pattern = "incoming")
ht1 + ht2

# Signaling role analysis on the cell-cell communication networks of interest
ht <- netAnalysis_signalingRole_heatmap(cellchat, signaling = c("CXCL", "CCL"))

Identify global communication patterns to explore how multiple cell types and signaling pathways coordinate together

In addition to exploring detailed communications for individual pathways, an important question is how multiple cell groups and signaling pathways coordinate to function. CellChat employs a pattern recognition method to identify the global communication patterns.

As the number of patterns increases, there might be redundant patterns, making it difficult to interpret the communication patterns. We chose five patterns as default. Generally, it is biologically meaningful with the number of patterns greater than 2. In addition, we also provide a function selectK to infer the number of patterns, which is based on two metrics that have been implemented in the NMF R package, including Cophenetic and Silhouette. Both metrics measure the stability for a particular number of patterns based on a hierarchical clustering of the consensus matrix. For a range of the number of patterns, a suitable number of patterns is the one at which Cophenetic and Silhouette values begin to drop suddenly.

Identify and visualize outgoing communication pattern of secreting cells

Outgoing patterns reveal how the sender cells (i.e. cells as signal source) coordinate with each other as well as how they coordinate with certain signaling pathways to drive communication.

To intuitively show the associations of latent patterns with cell groups and ligand-receptor pairs or signaling pathways, we used river (alluvial) plots. We first normalized each row of W and each column of H to be [0,1], and then set the elements in W and H to be zero if they are less than 0.5. Such thresholding allows to uncover the most enriched cell groups and signaling pathways associated with each inferred pattern, that is, each cell group or signaling pathway is associated with only one inferred pattern. These thresholded matrices W and H are used as inputs for creating alluvial plots.

To directly relate cell groups with their enriched signaling pathways, we set the elements in W and H to be zero if they are less than 1/R where R is the number of latent patterns. By using a less strict threshold, more enriched signaling pathways associated each cell group might be obtained. Using a contribution score of each cell group to each signaling pathway computed by multiplying W by H, we constructed a dot plot in which the dot size is proportion to the contribution score to show association between cell group and their enriched signaling pathways. USERS can also decrease the parameter cutoff to show more enriched signaling pathways associated each cell group.

Load required package for the communication pattern analysis

library(NMF)
#> Loading required package: pkgmaker
#> Loading required package: registry
#> Loading required package: rngtools
#> Loading required package: cluster
#> NMF - BioConductor layer [OK] | Shared memory capabilities [NO: bigmemory] | Cores 15/16
#>   To enable shared memory capabilities, try: install.extras('
#> NMF
#> ')
#> 
#> Attaching package: 'NMF'
#> The following objects are masked from 'package:igraph':
#> 
#>     algorithm, compare
library(ggalluvial)

Here we run selectK to infer the number of patterns.

selectK(cellchat, pattern = "outgoing")

Both Cophenetic and Silhouette values begin to drop suddenly when the number of outgoing patterns is 3.

nPatterns = 3
cellchat <- identifyCommunicationPatterns(cellchat, pattern = "outgoing", k = nPatterns)

# river plot
netAnalysis_river(cellchat, pattern = "outgoing")
#> Please make sure you have load `library(ggalluvial)` when running this function

# dot plot
netAnalysis_dot(cellchat, pattern = "outgoing")

Identify and visualize incoming communication pattern of target cells

Incoming patterns show how the target cells (i.e. cells as signal receivers) coordinate with each other as well as how they coordinate with certain signaling pathways to respond to incoming signals.

selectK(cellchat, pattern = "incoming")

Cophenetic values begin to drop when the number of incoming patterns is 4.

nPatterns = 4
cellchat <- identifyCommunicationPatterns(cellchat, pattern = "incoming", k = nPatterns)

# river plot
netAnalysis_river(cellchat, pattern = "incoming")
#> Please make sure you have load `library(ggalluvial)` when running this function

# dot plot
netAnalysis_dot(cellchat, pattern = "incoming")

Manifold and classification learning analysis of signaling networks

Further, CellChat is able to quantify the similarity between all significant signaling pathways and then group them based on their cellular communication network similarity. Grouping can be done either based on the functional or structural similarity.

Functional similarity: High degree of functional similarity indicates major senders and receivers are similar, and it can be interpreted as the two signaling pathways or two ligand-receptor pairs exhibit similar and/or redundant roles. The functional similarity analysis requires the same cell population composition between two datasets.

Structural similarity: A structural similarity was used to compare their signaling network structure, without considering the similarity of senders and receivers.

Identify signaling groups based on their functional similarity

cellchat <- computeNetSimilarity(cellchat, type = "functional")
cellchat <- netEmbedding(cellchat, type = "functional")
#> Manifold learning of the signaling networks for a single dataset
cellchat <- netClustering(cellchat, type = "functional")
#> Classification learning of the signaling networks for a single dataset
# Visualization in 2D-space
netVisual_embedding(cellchat, type = "functional", label.size = 3.5)

# netVisual_embeddingZoomIn(cellchat, type = "functional", nCol = 2)

Identify signaling groups based on structure similarity

cellchat <- computeNetSimilarity(cellchat, type = "structural")
cellchat <- netEmbedding(cellchat, type = "structural")
#> Manifold learning of the signaling networks for a single dataset
cellchat <- netClustering(cellchat, type = "structural")
#> Classification learning of the signaling networks for a single dataset
# Visualization in 2D-space
netVisual_embedding(cellchat, type = "structural", label.size = 3.5)

netVisual_embeddingZoomIn(cellchat, type = "structural", nCol = 2)

Part V: Save the CellChat object

saveRDS(cellchat, file = "cellchat_humanSkin_LS.rds")
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] parallel  stats     graphics  grDevices utils     datasets  methods  
#> [8] base     
#> 
#> other attached packages:
#>  [1] doParallel_1.0.16   iterators_1.0.13    foreach_1.5.1      
#>  [4] ggalluvial_0.12.3   NMF_0.23.0          cluster_2.1.2      
#>  [7] rngtools_1.5.2      pkgmaker_0.32.2     registry_0.5-1     
#> [10] patchwork_1.1.1     CellChat_1.6.0      Biobase_2.54.0     
#> [13] BiocGenerics_0.40.0 ggplot2_3.3.5       igraph_1.3.4       
#> [16] dplyr_1.0.7        
#> 
#> loaded via a namespace (and not attached):
#>   [1] utf8_1.2.2            reticulate_1.22       tidyselect_1.1.1     
#>   [4] htmlwidgets_1.5.4     grid_4.1.2            BiocParallel_1.28.3  
#>   [7] Rtsne_0.15            munsell_0.5.0         codetools_0.2-18     
#>  [10] ica_1.0-2             future_1.23.0         miniUI_0.1.1.1       
#>  [13] withr_2.4.3           colorspace_2.0-2      highr_0.9            
#>  [16] knitr_1.40            rstudioapi_0.13       Seurat_4.0.6         
#>  [19] stats4_4.1.2          ROCR_1.0-11           ggsignif_0.6.3       
#>  [22] tensor_1.5            listenv_0.8.0         labeling_0.4.2       
#>  [25] polyclip_1.10-0       farver_2.1.0          rprojroot_2.0.2      
#>  [28] coda_0.19-4           parallelly_1.30.0     vctrs_0.3.8          
#>  [31] generics_0.1.1        xfun_0.33             R6_2.5.1             
#>  [34] clue_0.3-60           spatstat.utils_2.3-0  assertthat_0.2.1     
#>  [37] promises_1.2.0.1      scales_1.1.1          gtable_0.3.0         
#>  [40] globals_0.14.0        goftest_1.2-3         rlang_0.4.12         
#>  [43] systemfonts_1.0.2     GlobalOptions_0.1.2   splines_4.1.2        
#>  [46] rstatix_0.7.0         lazyeval_0.2.2        spatstat.geom_2.3-1  
#>  [49] broom_0.7.10          yaml_2.2.1            reshape2_1.4.4       
#>  [52] abind_1.4-5           ggnetwork_0.5.10      backports_1.4.1      
#>  [55] httpuv_1.6.4          tools_4.1.2           gridBase_0.4-7       
#>  [58] statnet.common_4.5.0  ellipsis_0.3.2        spatstat.core_2.3-2  
#>  [61] jquerylib_0.1.4       RColorBrewer_1.1-2    ggridges_0.5.3       
#>  [64] Rcpp_1.0.7            plyr_1.8.6            purrr_0.3.4          
#>  [67] ggpubr_0.4.0          rpart_4.1-15          deldir_1.0-6         
#>  [70] pbapply_1.5-0         GetoptLong_1.0.5      cowplot_1.1.1        
#>  [73] S4Vectors_0.32.3      zoo_1.8-9             SeuratObject_4.0.4   
#>  [76] ggrepel_0.9.1         here_1.0.1            magrittr_2.0.1       
#>  [79] data.table_1.14.2     RSpectra_0.16-0       sna_2.6              
#>  [82] magick_2.7.3          scattermore_0.7       circlize_0.4.13      
#>  [85] lmtest_0.9-39         RANN_2.6.1            fitdistrplus_1.1-6   
#>  [88] matrixStats_0.61.0    mime_0.12             evaluate_0.17        
#>  [91] xtable_1.8-4          IRanges_2.28.0        gridExtra_2.3        
#>  [94] shape_1.4.6           compiler_4.1.2        tibble_3.1.6         
#>  [97] KernSmooth_2.23-20    crayon_1.4.2          htmltools_0.5.2      
#> [100] mgcv_1.8-38           later_1.3.0           tidyr_1.1.4          
#> [103] DBI_1.1.2             ComplexHeatmap_2.10.0 MASS_7.3-54          
#> [106] Matrix_1.3-4          car_3.0-12            cli_3.1.0            
#> [109] forcats_0.5.1         pkgconfig_2.0.3       plotly_4.10.0        
#> [112] spatstat.sparse_2.1-0 svglite_2.0.0         bslib_0.3.1          
#> [115] stringr_1.4.0         digest_0.6.29         sctransform_0.3.2    
#> [118] RcppAnnoy_0.0.19      spatstat.data_2.1-2   rmarkdown_2.17       
#> [121] leiden_0.3.9          uwot_0.1.11           shiny_1.7.1          
#> [124] rjson_0.2.20          lifecycle_1.0.1       nlme_3.1-153         
#> [127] jsonlite_1.7.2        carData_3.0-4         network_1.17.1       
#> [130] BiocNeighbors_1.12.0  viridisLite_0.4.0     fansi_0.5.0          
#> [133] pillar_1.6.4          lattice_0.20-45       fastmap_1.1.0        
#> [136] httr_1.4.2            survival_3.2-13       glue_1.6.0           
#> [139] FNN_1.1.3             png_0.1-7             stringi_1.7.6        
#> [142] sass_0.4.0            irlba_2.3.5           future.apply_1.8.1