scDblFinder去除双胞出错

Question

求./助，我首先将单细胞数据进行标准化和降维处理，然后用scDblFinder去除双胞，但是一直报错。

object <- Read10X('E:\\R\\brain\\MER2\\10K3') #读取单细胞数据
object <- CreateSeuratObject(counts = object, project = "brain_1", min.cells = 3, min.features = 200) #创建seurat对象
object[["percent.mt"]] <- PercentageFeatureSet(object, pattern = "ATP6|ATP8|COX1|COX2|COX3|CYTB|ND1|ND2|ND3|ND4|ND4L|ND5|ND6")
if (sum([e-mail]$percent.mt) == 0){
  [e-mail]$percent.mt[1] <- 0.000001
} ## 计算线粒体基因百分比，0值补成非0的极小数
object <- subset(object, subset = nFeature_RNA > 200 & nFeature_RNA < 6000 & nCount_RNA < 20000 & percent.mt < 5) # 200 < nFeature_RNA(单个细胞总基因数) < 6000; # nCount_RNA(单个细胞总表达量数) < 20000; # percent.mt(单个细胞内线粒体基因比例) < 5
# 进行seurat流程的预处理，三步标准化。
object <- NormalizeData(object)
object <- FindVariableFeatures(object, selection.method = "vst", nfeatures = 2000)
object <- ScaleData(object)
object <- RunPCA(object)
object <- RunUMAP(object, dims = 1:20)
object <- FindNeighbors(object, reduction = "pca", dims = 1:20)
object <- FindClusters(object, resolution = 0.5)

sce <- as.SingleCellExperiment(object)

sce <- scDblFinder(sce, dbr=0.1)

Answer 1

让【宝码香车】来帮你解答，本回答参考DeepSeek编写，并整理提供。
如果还有疑问可以点击头像关注私信或评论。如果答案让您满意，请采纳、点赞、关注，非常感谢！

根据您提供的错误信息和代码，问题出在scDblFinder函数处理DelayedMatrix对象时无法正确提取数组。以下是解决方案：

1. 问题原因

错误信息表明scDblFinder无法处理DelayedMatrix类型的对象，这是由SingleCellExperiment转换过程中保留的延迟操作导致的。

2. 解决方案

在转换为SingleCellExperiment之前，先将数据转换为常规矩阵：

# 修改后的代码
sce <- as.SingleCellExperiment(object)
# 将counts转换为常规矩阵
counts(sce) <- as.matrix(counts(sce))
# 运行scDblFinder
sce <- scDblFinder(sce, dbr=0.1)

3. 完整修正代码

# 读取和处理数据
object <- Read10X('E:\\R\\brain\\MER2\\10K3')
object <- CreateSeuratObject(counts = object, project = "brain_1", min.cells = 3, min.features = 200)
object[["percent.mt"]] <- PercentageFeatureSet(object, pattern = "ATP6|ATP8|COX1|COX2|COX3|CYTB|ND1|ND2|ND3|ND4|ND4L|ND5|ND6")
if (sum([e-mail]$percent.mt) == 0){
  [e-mail]$percent.mt[1] <- 0.000001
}
object <- subset(object, subset = nFeature_RNA > 200 & nFeature_RNA < 6000 & nCount_RNA < 20000 & percent.mt < 5)

# 标准化流程
object <- NormalizeData(object)
object <- FindVariableFeatures(object, selection.method = "vst", nfeatures = 2000)
object <- ScaleData(object)
object <- RunPCA(object)
object <- RunUMAP(object, dims = 1:20)
object <- FindNeighbors(object, reduction = "pca", dims = 1:20)
object <- FindClusters(object, resolution = 0.5)

# 转换并修复DelayedMatrix问题
sce <- as.SingleCellExperiment(object)
counts(sce) <- as.matrix(counts(sce))
sce <- scDblFinder(sce, dbr=0.1)

4. 其他建议

1. 如果数据量很大，转换矩阵可能消耗较多内存，可以考虑：

   library(DelayedArray)
   setAutoBlockSize(1e6) # 调整块大小
   

2. 确保所有包都是最新版本：

   update.packages(ask = FALSE, checkBuilt = TRUE)
   

3. 如果仍然报错，可以尝试先转换为稀疏矩阵：

   counts(sce) <- as(counts(sce), "dgCMatrix")