Welcome to STcomm, an R package to illustrate the spatially resolved cell interactions by combined the spatial cellular colocalization with their enriched ligand-receptor co-expression patterns inferred from both spatial and single-cell transcriptomic data.
STcomm R package can be easily installed from Github using devtools:
devtools::install_github("Vanjia-lee/STcomm")Firstly, rctd4weights function will help you tidy the confident
weights for every pixel after RCTD analysis.
library(STcomm)
rctd.multi <- rctd4weight(rctd_obj, rctd_mode = 'multi', conf = TRUE)Secondly, cellColocation function will help you quantify the colocalization
of cell type pairs within spots by calculating Pearson correlation coefficient
or Jaccard similarity coefficient based on cell type composition predicted by
RCTD.
# By calculating Pearson correlation coefficient based on the RCTD object
colocal_ctps1 <- cellColocation(rctd_obj, rctd_mode = 'multi',
method = 'pcc', pcc = 0.06,
pval = 0.05, padj = 0.05)
# By calculating Pearson correlation coefficient based on the data.frame of RCTD
confident weights (with conf = TRUE when RCTD) for each pixel
colocal_ctps2 <- cellColocation(rctd_df, method = 'pcc',
pcc = 0.06, pval = 0.05, padj = 0.05)
# By calculating Jaccard similarity coefficient based on the RCTD object
colocal_ctps3 <- cellColocation(rctd_obj, rctd_mode = 'multi',
method = 'jac', jac = 0.05)
# By calculating Jaccard similarity coefficient based on the data.frame of RCTD
confident weights (with conf = TRUE when RCTD) for each pixel
colocal_ctps4 <- cellColocation(rctd_df, method = 'jac', jac = 0.05)Then, you can identify significant co-occurrence cell type groups belonging to the same spot from the cell type colocalization network.
Thirdly, based on the spatial data, you can obtain significantly co-expressed
Ligand-Receptor (LR) pairs for spatially co-localized cell types by performing
Fisher's exact test on binarized co-localized cell type pairs and co-expressed
LR pairs at spot level. Next, you calculate significant communication between
LR pairs in co-localized cell type pairs based on the refrence single cell
transcriptomic data. Finally, to get high confidence and spatial aware cell-cell
communication, you can keep only spatially relevant communication information
based on the above Fisher exact significancy. st_comm function can help you
characterize confident spatially resolved cell-cell interaction with the tissue
organization.
# load the CellChat object for the refrence single cell transcriptomic data
cellchat <- readRDS(cellchat)
st_net <- st_comm(st_obj, weights.df = rctd_multi, ctpairs = colocal_ctps1, cellchat = cellchat)
# or you would like to prepare a data.frame frome the CellChat object, which
# contained the communication probability and infered cellular communication network.
net.df <- subsetCommunication(cellchat)
net.df$ct_pairs <- paste0(net.df$source, "_", net.df$target)
st_net <- st_comm(st_obj, weights.df = rctd_multi, ctpairs = colocal_ctps1, cellchat = net.df)