Seurat: Cell hashing

1. 官网教程

https://satijalab.org/seurat/articles/hashing_vignette

2. 博文教程

• https://www.cn-healthcare.com/articlewm/20210831/content-1259008.html

• https://www.jianshu.com/p/d5aaecc76214

• https://www.jianshu.com/p/4d0018b8a52a

• https://cloud.tencent.com/developer/article/1680657

3. HTODemux()

https://satijalab.org/seurat/articles/hashing_vignette.html

Seurat_object_HTO <- HTODemux(Seurat_object_HTO, assay="HTO", positive.quantile=0.8)

positive.quantile=: 用来区分 hashtag 信号的强度；高于此数值则为 positive；默认值为 0.99

输出结果：

HTO_maxID: Name of hashtag with the highest signal

HTO_secondID: Name of hashtag with the second highest signal

HTO_margin: The difference between signals for hash.maxID and hash.secondID

HTO_classification: Classification result, with doublets/multiplets named by the top two highest hashtags

HTO_classification.global: Global classification result (singlet, doublet or negative)

hash.ID: Classification result where doublet IDs are collapsed

• 其中，HTO_classification, HTO_classification.global 和 hash.ID 三列的结果是对等的

• 根据需求和标准不同，HTO_maxID 和 hash.ID 均可作为分类依据

4. MULTIseqDemux()

https://satijalab.org/seurat/reference/multiseqdemux

Seurat_object_HTO <- MULTIseqDemux(Seurat_object_HTO, assay="HTO", quantile=0.7)

positive.quantile=: 用来区分 hashtag 信号的强度；高于此数值则为 positive；默认值为 0.7

Seurat 中另一个用来根据 hashtag 信号进行细胞分类的函数

输出结果：

MULTI_ID: Classification result where doublet IDs are collapsed

MULTI_classification: Classification result, with doublets/multiplets named by the top two highest hashtags

• 两个结果是对等的，可以根据需求来选择用哪个结果作为分类标准

5. 基本流程

require(Seurat)

Seurat_object_HTO <- NormalizeData(Seurat_object_HTO, assay = "HTO", normalization.method = "CLR")   '#用 CLR 方法对 HTO 信号进行归一化

Seurat_object_HTO <- HTODemux(Seurat_object_HTO, assay = "HTO", positive.quantile = 0.8)   #细胞分类方式一

Seurat_object_HTO <- MULTIseqDemux(Seurat_object_HTO, assay = "HTO")   #细胞分类方式二

6. 结果展示

https://satijalab.org/seurat/articles/hashing_vignette.html

'#Calculate the tSNE embedding based on the HTO matrix
DefaultAssay(sample_mix) <- "HTO"
sample_mix <- ScaleData(sample_mix, features = rownames(sample_mix),verbose = FALSE)
sample_mix <- RunPCA(sample_mix, features = rownames(sample_mix), approx = FALSE)
sample_mix <- RunTSNE(sample_mix, dims = 1:5, perplexity = 100, reduction = "pca", check_duplicates=F)
DimPlot(sample_mix)

#----------------------------------------------------------------------------------------#
#The tSNE dimensional reduction based on the PC of HTO matrix
DimPlot(sample_mix)

#----------------------------------------------------------------------------------------#
#Heatmap shows the signal variance of different cells
HTOHeatmap(sample_mix, assay = "HTO")