文章目录
brief
- seurat提供的教学里面包含了
Standard pre-processing workflow,workflow包括QC,normalization,scale data ,detection of highly variable features。 - 其中 normalization就有蛮多方法的,seurat自己就提供了两种,一种是"LogNormalization",另一种是目前正在大力推荐的"SCTransform",而且每一种方法参数众多,我觉得有必要对其进行仔细的探究,所以需要分开学习和记录。
- 还有 detection of highly variable features以及scale data,也是方法和参数众多,我觉得有必要对其进行仔细的探究,所以需要分开学习和记录
- 概要图以及系列博文可以参见链接。

这里主要记录了Normalize Data的学习过程,包括LogNormalization 和 SCTransform两种normalization方法的参数以及对数据的影响,以及对下游数据处理流程的影响等。
为什么要做normalization
scRNA-seq中细胞与细胞之间的异质性可能是生物学上的,也可能是技术上造成的。其中每个细胞检测到的分子数,测序深度对细胞之间的变异度贡献很大,需要进行矫正。



实例
Normalization之前的预处理
library(dplyr)
library(Seurat)
library(patchwork)
rm(list=ls())
# 使用read10X读取output of the cellranger pipeline from 10X,包括barcodes,genes,matrix.mtx三个文件
pbmc.data <- Read10X(data.dir = "D:/djs/pbmc3k_filtered_gene_bc_matrices/filtered_gene_bc_matrices/hg19")
# 使用 CreateSeuratObject函数构造seurat对象
pbmc <- CreateSeuratObject(counts = pbmc.data, project = "pbmc3k",
min.cells = 3, min.features = 200,
names.delim = "-",names.field = 1)
pbmc[["percent.mt"]] <- PercentageFeatureSet(pbmc, pattern = "^MT-")
pbmc <- subset(pbmc, subset = nFeature_RNA > 200 & nFeature_RNA < 2500 & percent.mt < 5)
NormalizeData --> LogNormalize

-
object
上述构建的seurat object -
normalization.method
-
LogNormalize: Feature counts for each cell are divided by the total counts for that cell and multiplied by the scale.factor. This is then natural-log transformed using log1p.For counts per million (CPM) set scale.factor = 1e6
-
CLR: Applies a centered log ratio transformation
-
RC: Relative counts. Feature counts for each cell are divided by the total counts for that cell and multiplied by the scale.factor. No log-transformation is applied.
-
-
scale.factor
Sets the scale factor for cell-level normalization。For counts per mil
264



被折叠的 条评论
为什么被折叠?



