利用MetaNeighbor验证重复性和跨物种分群

进行跨物种研究时，我们经常需要进行注释结果的比较和归类，或者同一物种不同样本之间的注释验证。R语言中有一个包就可以利用直观的热图展示这一需求。

导入包和环境

library(Seurat)
library(ggplot2)
library(MetaNeighbor)
library(SingleCellExperiment)
library(dplyr)

导入数据

这里以海豚和两个公开人的PFC脑区单细胞转录组数据为例

hm <- readRDS('human_sci_EXN.rds')
DefaultAssay(hm) <- 'RNA'
[email protected]$species <- 'human'
hm2 <- readRDS('human_2_EXN.rds')
DefaultAssay(hm2) <- 'RNA'
[email protected]$species <- 'human'
dp <- readRDS('dolphin_PFC_EXN.rds')
DefaultAssay(dp) <- 'RNA'
[email protected]$species <- 'dolphin'

创建比较对象

# 确保Idents已经正确设置
Idents(hm) <- [email protected]$cellType_layer
Idents(hm2) <- [email protected]$subclass.v2
Idents(dp) <- [email protected]$sub2

# 提取表达矩阵和元数据
exprs_hm <- GetAssayData(hm)
exprs_hm2 <- GetAssayData(hm2)
exprs_dp <- GetAssayData(dp)

meta_hm <- [email protected]
meta_hm2 <- [email protected]
meta_dp <- [email protected]

# 提取共有基因
common_genes <- Reduce(intersect, list(rownames(exprs_hm),rownames(exprs_hm2), rownames(exprs_dp)))

# 提取共有基因表达矩阵
exprs_hm_common <- exprs_hm[common_genes, ]
exprs_hm2_common <- exprs_hm2[common_genes, ]
exprs_dp_common <- exprs_dp[common_genes, ]

# 合并表达矩阵
combined_exprs <- cbind(exprs_hm_common, exprs_hm2_common, exprs_dp_common)

# 创建一个新的列数据框，确保使用Idents作为cell_type
new_colData <- data.frame(
  study_id = c(rep('human', ncol(exprs_hm_common)),rep('human2', ncol(exprs_hm2_common)), rep('dolphin', ncol(exprs_dp_common))),
  cell_type = c(Idents(hm), Idents(hm2), Idents(dp))
)

# 创建 SingleCellExperiment 对象
sce_combined <- SingleCellExperiment(assays = list(RNA = combined_exprs), colData = new_colData)

# 检查合并后的对象
sce_combined

检查所有细胞分群比例

table(sce_combined$cell_type)

相似性分析

# 选择高可变基因
var_genes <- variableGenes(dat = sce_combined, exp_labels = sce_combined$study_id)

# 运行相似性分析
celltype_NV <- MetaNeighborUS(var_genes = var_genes,
                              dat = sce_combined,
                              study_id = sce_combined$study_id,
                              cell_type = sce_combined$cell_type,
                              fast_version = TRUE)
# 检查前几行
head(celltype_NV)

可视化

png("heatmap_integrated_3.png", width = 24, height = 24, units = "in", res = 300) 
plotHeatmapPretrained(aurocs = celltype_NV, cex = 1.5, margins = c(25, 25))
dev.off()