单细胞转录组数据分析（二）：聚类clustering

整合scRNA数据后，那么就是聚类处理，选择合适的聚类参数是识别celltype的重点。更多知识分享请到 https://zouhua.top/。

加载R包

library(dplyr)
library(Seurat)
library(ggplot2)

# 设置参数
rm(list = ls())
options(stringsAsFactors = F)
options(future.globals.maxSize = 4000 * 1024^2)

colors <- c("#A6CEE3", "#1F78B4", "#08306B", "#B2DF8A", "#006D2C", "#8E0152",
         "#DE77AE", "#CAB2D6", "#6A3D9A", "#FB9A99", "#E31A1C", "#B15928",
         "#619CFF","#FF67A4","#00BCD8", "#EE2B2B", "#2D6BB4")

读取数据

data.integrated <- readRDS("../../Result/RDS/data.n8.sct.rds")

聚类前可视化

# 若原数据没有PCA结果，则运行下面部分
data.integrated <- RunPCA(object = data.integrated, verbose = FALSE)
data.integrated <- RunTSNE(data.integrated, reduction = "pca", dims = 1:30)
data.integrated <- RunUMAP(data.integrated, reduction = "pca", dims = 1:30)

cowplot::plot_grid(
        DimPlot(data.integrated,
        reduction = "umap",
        label = TRUE,
        label.size = 6),
        DimPlot(data.integrated,
        reduction = "umap",
        label = TRUE,
        label.size = 6,
        split.by = "Batch")+NoLegend(),
        ncol = 2)

查看PC的metagenes(热图展示)

DimHeatmap(data.integrated,
           reduction = "pca",
           dims = 1:9,
           cells = 500,
           balanced = TRUE)

print(x = data.integrated[["pca"]],
      dims = 1:10,
      nfeatures = 5)

ElbowPlot(object = data.integrated,
          ndims = 40)+
        geom_vline(xintercept = 30, linetype=2)+
        geom_hline(yintercept = 2, linetype=2)

聚类，设置不同的分辨率

# Determine the K-nearest neighbor graph
data.integrated <- FindNeighbors(object = data.integrated,
                                dims = 1:30)

# Determine the clusters for various resolutions
data.integrated <- FindClusters(object = data.integrated,
                               resolution = c(0.4, 0.6, 0.8, 1.0, 1.4))

查看不同分辨率下簇的分布情况

• 簇之间明显区分；

• 簇没有跨区域分散；

Idents(object = data.integrated) <- "integrated_snn_res.0.4"
DimPlot(data.integrated,
        reduction = "umap",
        label = TRUE,
        label.size = 6)

Idents(object = data.integrated) <- "integrated_snn_res.0.6"
DimPlot(data.integrated,
        reduction = "umap",
        label = TRUE,
        label.size = 8)

Idents(object = data.integrated) <- "integrated_snn_res.1.4"
DimPlot(data.integrated,
        reduction = "umap",
        label = TRUE,
        label.size = 6)

注意：如没有integration步骤，则使用RNA_snn_res.0.6。

根据聚类结果选择resolution参数

Idents(object = data.integrated) <- "integrated_snn_res.0.4"
saveRDS(data.integrated, "../../Result/RDS/data.all.cluster04.rds", compress = TRUE)

Reference

参考文章如引起任何侵权问题，可以与我联系，谢谢。