单细胞转录组Seurat去批次-RPCA算法

探序基因已入锐竞平台

要将每个样本的矩阵都单独建立Seurat对象。

如果所有样本在一个Seurat对象中，并且在meta表格中有一列记载细胞对应的样本名，可按如下方法切分：

复制代码

# split the dataset into a list of two seurat objects (stim and CTRL)
ifnb.list <- SplitObject(ifnb, split.by = "stim")

# normalize and identify variable features for each dataset independently
ifnb.list <- lapply(X = ifnb.list, FUN = function(x) {
    x <- NormalizeData(x)
    x <- FindVariableFeatures(x, selection.method = "vst", nfeatures = 2000)
})

或者：

例如在Seurat对象中，meta表格的Samples记录了细胞对应的样本名，运行如下代码切分

project.Seurat $\["RNA"$ ] <- split(project.Seurat $\["RNA"$ ], f = project.Seurat$Samples)

最后count矩阵会按样本切分：

> project.Seurat $\["RNA"$ ]

Assay (v5) data with xxx features for xxx cells

First 10 features:

xxx

Layers:

counts.S1, counts.S2, counts.S3, counts.S4

切分完后：

project.Seurat <- NormalizeData(project.Seurat)

project.Seurat <- FindVariableFeatures(project.Seurat)

project.Seurat <- ScaleData(project.Seurat)

project.Seurat <- RunPCA(project.Seurat)

project.Seurat <- IntegrateLayers(

object = project.Seurat, method = RPCAIntegration,

orig.reduction = "pca", new.reduction = "rpca",

verbose = FALSE

)

大细胞量评测：

细胞量：大约90万个，Seurat版本：5.5.0，SeuratObject版本：5.4.0

报错：

Warning in svd.function(A = t(x = object), nv = npcs, ...) :

You're computing too large a percentage of total singular values, use a standard svd instead.

Warning in svd.function(A = t(x = object), nv = npcs, ...) :

You're computing too large a percentage of total singular values, use a standard svd instead.

Warning in svd.function(A = t(x = object), nv = npcs, ...) :

You're computing too large a percentage of total singular values, use a standard svd instead.

Error in idx $i,$ <- res $\[i$ ] $\[1$ ] : 被替换的项目不是替换值长度的倍数

Calls: IntegrateLayers ... resolve.list -> signalConditionsASAP -> signalConditions

停止执行

参考：

Seurat Tutorial 5：使用 reciprocal PCA (RPCA) 快速整合https://zhuanlan.zhihu.com/p/653865719 Seurat4.0系列教程13：使用RPCA快速整合数据https://cloud.tencent.com/developer/article/1931214 单细胞：IntegrateLayers函数中这几种主流的单细胞数据整合方法https://cloud.tencent.com/developer/article/2616679