单细胞分析之CellChat

来源：单细胞天地评论2,755

CellChat简介

CellChat的特点

  1. 全面的数据：大多数的细胞通讯分析方法通常只考虑配体/受体基因对，往往忽略了多亚基复合物受体和其他信号辅助因子。CellChat的作者人工挑选了2021个经过验证的细胞通讯关系，构建了新的细胞通讯参考数据库——CellChatDB。它不仅考虑了多亚基受体情况，还收录了其他重要的信号辅助因子：可溶性激动剂，拮抗剂，共刺激和共抑制膜结合受体。
  2. 高深的算法：CellChat在分析过程中使用了多种分析方法，不仅有常见的秩和检验、置换检验、SNN、KNN、UMAP，还有非负矩阵分解、社会网络分析、质量作用定律等不常用的方法。能灵活运用这么多方法分析数据，足以说明作者具有深厚的算法功底。
  3. 丰富的可视化结果：CellChat提供丰富且美观的可视化结果，有网络图、桑基图、热图、气泡图、散点图等多种图形。

安装CellChat

安装依赖项

  1. ## R环境安装NMF和ComplexHeatmap包
  2. devtools::install_github("renozao/NMF@devel")
  3. devtools::install_github("jokergoo/ComplexHeatmap")
  4.  
  5. ## shell环境安装
  6. pip install umap-learn  #如果报错，可使用conda安装
  7. #conda install umap-learn

安装CellChat

  1. devtools::install_github("sqjin/CellChat")

CellChat实践

数据来源

为了方便大家对比不同的分析方法，我们继续使用《单细胞分析之NicheNet》的数据。数据下载链接：https://zenodo.org/record/3531889/files/seuratObj.rds

初始分析

  1. library(CellChat)
  2. library(tidyverse)
  3. library(ggalluvial)
  4. rm(list=ls())
  5. options(stringsAsFactors = FALSE)
  6.  
  7. ##提取表达矩阵和细胞分类信息
  8. scRNA <- readRDS(url("https://zenodo.org/record/3531889/files/seuratObj.rds"))
  9. scRNA <- UpdateSeuratObject(scRNA)
  10. # CellChat要求输入标准化后的表达数据
  11. data.input <- GetAssayData(scRNA, assay = "RNA", slot = "data")
  12. identity <- subset(scRNA@meta.data, select = "celltype")
  13.  
  14. ##创建cellchat对象
  15. cellchat <- createCellChat(data = data.input)
  16. cellchat <- addMeta(cellchat, meta = identity, meta.name = "labels")
  17. cellchat <- setIdent(cellchat, ident.use = "labels") 
  18. groupSize <- as.numeric(table(cellchat@idents)) # 后面有用
  19.  
  20. ##设置参考数据库
  21. # 选择合适的物种，可选CellChatDB.human, CellChatDB.mouse
  22. CellChatDB <- CellChatDB.mouse  
  23. # 使用"Secreted Signaling"用于细胞通讯分析
  24. CellChatDB.use <- subsetDB(CellChatDB, search = "Secreted Signaling") 
  25. # 将数据库传递给cellchat对象
  26. cellchat@DB <- CellChatDB.use 
  27.  
  28. ##配体-受体分析
  29. # 提取数据库支持的数据子集
  30. cellchat <- subsetData(cellchat)
  31. # 识别过表达基因
  32. cellchat <- identifyOverExpressedGenes(cellchat)
  33. # 识别配体-受体对
  34. cellchat <- identifyOverExpressedInteractions(cellchat)
  35. # 将配体、受体投射到PPI网络
  36. cellchat <- projectData(cellchat, PPI.mouse)

推断细胞通讯网络

  1. ##推测细胞通讯网络
  2. cellchat <- computeCommunProb(cellchat)
  3. cellchat <- computeCommunProbPathway(cellchat)
  4. cellchat <- aggregateNet(cellchat)

细胞通讯网络系统分析及可视化

  1. levels(cellchat@idents)            #查看细胞顺序
  2. vertex.receiver = c(3, 6)          #指定靶细胞的索引
  3. cellchat@netP$pathways             #查看富集到的信号通路
  4. pathways.show <- "CCL"             #指定需要展示的通路
  5. # Hierarchy plot
  6. png(filename = "sig_pathway_hierarchy.png", width = 1000, height = 650)
  7. netVisual_aggregate(cellchat, signaling = pathways.show,  vertex.receiver = vertex.receiver, vertex.size = groupSize)
  8. dev.off()
  9. # Circle plot
  10. png(filename = "sig_pathway_cricle.png", width = 650, height = 600)
  11. netVisual_aggregate(cellchat, signaling = pathways.show, layout = "circle", vertex.size = groupSize)
  12. dev.off()
  13. # 计算配体-受体对信号网络的贡献度
  14. png(filename = "sig_pathway_L-R.png", width = 800, height = 600)
  15. netAnalysis_contribution(cellchat, signaling = pathways.show)
  16. dev.off()
  17. # 分析细胞在信号网络中角色
  18. cellchat <- netAnalysis_signalingRole(cellchat, slot.name = "netP")
  19. png(filename = "sig_pathway_role.png", width = 800, height = 600)
  20. netVisual_signalingRole(cellchat, signaling = pathways.show)
  21. dev.off()

sig_pathway_hierarchy

Image

sig_pathway_cricle

Image

sig_pathway_L-R

Image

sig_pathway_role

Image
  1. ##细胞通讯模式和信号网络
  2. nPatterns = 5   #默认为5
  3. cellchat <- identifyCommunicationPatterns(cellchat, pattern = "outgoing", k = nPatterns)
  4. # river plot
  5. p = netAnalysis_river(cellchat, pattern = "outgoing")
  6. ggsave("com_pattern_outgoing_river.png", p, width = 12, height = 6)
  7. # dot plot
  8. p = netAnalysis_dot(cellchat, pattern = "outgoing")
  9. ggsave("com_pattern_outgoing_dot.png", p, width = 9, height = 6)
  10.  
  11. nPatterns = 5
  12. cellchat <- identifyCommunicationPatterns(cellchat, pattern = "incoming", k = nPatterns)
  13. # river plot
  14. p = netAnalysis_river(cellchat, pattern = "incoming")
  15. ggsave("com_pattern_incoming_river.png", p, width = 12, height = 6)
  16. # dot plot
  17. p = netAnalysis_dot(cellchat, pattern = "incoming")
  18. ggsave("com_pattern_incoming_dot.png", p, width = 9, height = 6)

com_pattern_outgoing_river

Image

com_pattern_outgoing_dot

Image

com_pattern_incoming_river

Image

om_pattern_incoming_dot

Image
  1. ##信号网络聚类
  2. # 按功能相似性聚类
  3. cellchat <- computeNetSimilarity(cellchat, type = "functional")
  4. cellchat <- netEmbedding(cellchat, type = "functional")
  5. cellchat <- netClustering(cellchat, type = "functional")
  6. # Visualization in 2D-space
  7. p = netVisual_embedding(cellchat, type = "functional")
  8. ggsave("custer_pathway_function.png", p, width = 9, height = 6)
  9. p = netVisual_embeddingZoomIn(cellchat, type = "functional")
  10. ggsave("custer_pathway_function2.png", p, width = 8, height = 6)
  11.  
  12. # 按结构相似性聚类
  13. cellchat <- computeNetSimilarity(cellchat, type = "structural")
  14. cellchat <- netEmbedding(cellchat, type = "structural")
  15. cellchat <- netClustering(cellchat, type = "structural")
  16. # Visualization in 2D-space
  17. p = netVisual_embedding(cellchat, type = "structural")
  18. ggsave("custer_pathway_structure.png", p, width = 9, height = 6)
  19. p = netVisual_embeddingZoomIn(cellchat, type = "structural")
  20. ggsave("custer_pathway_structure2.png", p, width = 8, height = 6)
  21.  
  22. save(cellchat, file = "cellchat.rds")

custer_pathway_function

Image

custer_pathway_structure

Image

发表评论

匿名网友

拖动滑块以完成验证
加载中...