核心概念
一、細胞通訊
在組織微環境中，細胞透過分泌配體 (Ligand)與鄰近細胞表面的受體 (Receptor)結合，觸發下游訊號傳導。scRNA-seq 能同時測量所有細胞的基因表達，因此可以推斷哪些細胞在「發送」什麼訊號、哪些在「接收」。
In tissue microenvironments, cells communicate by secreting ligands that bind to receptors on neighboring cells, triggering downstream signaling. scRNA-seq measures all cells simultaneously, enabling inference of who is "sending" and who is "receiving."
互動模擬
二、網絡探索器
選擇一個「發送方」細胞類型，觀察與各接收方之間的通訊強度和代表性 L-R pairs。
Select a "sender" cell type to view communication strength with each receiver and representative L-R pairs.
主要 L-R pairs
方法比較
三、工具比較
| 工具 | 邏輯 | 特色 | ||
|---|---|---|---|---|
| CellChat (R) | 質量作用定律建模 L-R 交互機率 | 豐富視覺化；支援跨條件比較 | Mass action law for L-R interaction probability | Rich visualization; cross-condition comparison |
| CellPhoneDB (Py) | 排列檢定的 L-R 配對篩選 | 支援多亞基複合體 | Permutation-based L-R pair screening | Supports multi-subunit complexes |
| NicheNet (R) | 從配體追蹤到下游調控基因 | 不只看 L-R 配對，還追蹤下游效應 | Traces ligand to downstream target genes | Goes beyond L-R pairs to downstream effects |
| LIANA (R/Py) | 整合多種方法的統一框架 | 同時跑多種演算法取共識 | Unified framework integrating multiple methods | Run multiple algorithms, take consensus |
程式碼
四、實作範例
library(CellChat) cellchat <- createCellChat(pbmc, group.by = "cell_type") CellChatDB <- CellChatDB.human cellchat@DB <- CellChatDB cellchat <- subsetData(cellchat) cellchat <- identifyOverExpressedGenes(cellchat) cellchat <- identifyOverExpressedInteractions(cellchat) cellchat <- computeCommunProb(cellchat) cellchat <- computeCommunProbPathway(cellchat) cellchat <- aggregateNet(cellchat) netVisual_circle(cellchat@net$count) netVisual_bubble(cellchat, sources.use = 1, targets.use = c(2,3,4))
from cellphonedb.src.core.methods import cpdb_analysis_method results = cpdb_analysis_method.call( cpdb_file_path="cellphonedb.zip", meta_file_path="meta.txt", counts_file_path="counts.h5ad", counts_data="hgnc_symbol", output_path="./output")
重要提醒：這些分析都是「推斷」，不是直接證據。scRNA-seq 只能告訴我們 L-R 的 RNA 是否表達，但 RNA ≠ 蛋白質 ≠ 真正的物理交互。重要結果務必用實驗（共培養、阻斷實驗）驗證。Important: These analyses are inferences, not direct evidence. scRNA-seq only tells us if L-R RNAs are expressed — RNA ≠ protein ≠ actual physical interaction. Important findings must be validated experimentally (co-culture, blocking experiments).