cellchat.R

library(CellChat)
library(patchwork)
library(Seurat)

data.dir <- '13-cellchat'
dir.create(data.dir)
setwd(data.dir)
sce <- readRDS("../3-auto_annotate/sce.clean.rds")

meta2 <- read.delim("../../../meta.txt")
meta2 <- meta2[meta2$class1 == "scRNA",]
meta2$group <- meta2$group1

group1 <- unique(meta2$group)
load("../4-subtype2/finaltype/sce_meta.Rdata")

library(future)
options(future.seed = TRUE)
options(future.globals.maxSize = 10 * 1024^3)  # 2GB
plan("multisession", workers = 4) # 并行处理

a <- sum(rownames(sce_meta) != rownames(sce@meta.data))
if (a > 0){
  print("something was wrong")
}
meta <- sce_meta
meta <- meta %>% dplyr::select(orig.ident,celltype)
#meta3 <- meta %>% left_join(meta2, by = c("orig.ident" = "id"))
meta$group <- meta2$group[match(meta$orig.ident,meta2$id)]
chat1 <- c("celltype2","celltype3")
for (j in chat1[c(2)]) {
  unlink(j,recursive = T)
  dir.create(j)
  setwd(j)
  z <- match(rownames(meta),rownames(sce@meta.data))
  length(unique(z))
  sce$celltype2 <- sce$celltype
  sce$celltype2[match(rownames(sce_meta),rownames(sce@meta.data))] <- sce_meta[,j]
  for (i in group1) {
    cell.use = rownames(meta)[meta$group == i] #
    # B. 为 CelChat 分析子集输入数据
    sce.use <- sce[,cell.use]
    sce.use <- sce.use[,sce.use$celltype2 != "contamination"]
    sce.use$samples <- as.factor(sce.use$sample)
    cellchat <- createCellChat(object = sce.use, group.by = "celltype2",assay = "RNA")  #从seruat对象直接构建更方便。
    
    #每组数据需独立运行以下步骤：
    #(1) 配体-受体数据库预处理
    CellChatDB <- CellChatDB.human# 如果分析小鼠数据，请使用 CellChatDB.mouse
    showDatabaseCategory(CellChatDB)
    #CellChatDB.use <- subsetDB(CellChatDB, search = "Secreted Signaling", key = "annotation")# 使用分泌信号
    cellchat@DB <- CellChatDB
    #(2) 计算细胞通讯概率
    # 使用 CellChatDB 的一个子集进行细胞间通讯分析
    cellchat <- subsetData(cellchat)  # 过滤低表达基因
    cellchat <- identifyOverExpressedGenes(cellchat)
    cellchat <- identifyOverExpressedInteractions(cellchat)
    
    cellchat <- computeCommunProb(cellchat, type = "triMean")# 计算通讯概率
    #(3) 整合通路并过滤
    cellchat <- filterCommunication(cellchat, min.cells = 10)
    #用户可以过滤掉某些细胞群中细胞数量较少的细胞间通讯。默认情况下，每个细胞群中用于细胞间通讯的最小细胞数量为 10。
    ####在信号通路水平推断细胞间通讯####
    cellchat <- computeCommunProbPathway(cellchat)
    #计算聚合的细胞间通讯网络
    cellchat <- aggregateNet(cellchat)
    saveRDS(cellchat, file =  paste0(i,".rds"))
  }
  setwd("../")
}