I've made available some publicly available data containing single cell RNA-Seq data for 5k immune cells as well as a script that runs the Seurat workflow to define cell-type clusters in this data.
You can get the data, R script and results from ls6 here:
/work2/projects/BioITeam/projects/courses/rnaseq_course/day_5_single_cell_data Script: seuratTotalScript5k.mod.R Data (after preprocessing by cellRanger): filtered_feature_bc_matrix Results generated: results #You can copy over the data to your directory: cds cd my_rnaseq_course cp -r /work2/projects/BioITeam/projects/courses/rnaseq_course/day_5_single_cell_data . |
The script we are going to run is an older version (Because Seurat on TACC is an older version). You can kick it off by doing the following:
module load seurat-scripts/ctr-0.0.5--r341_0 #OPEN AN IDEV SESSION: idev -m 120 -q normal -A OTH21164 -r rna-seq-class-0613 R CMD BATCH seuratTotalScript5k.mod.3.4.R & |
The script does the following (newer versions of the functions given here):
Load required libraries
library(Seurat) library(ggplot2) library(dplyr) library(data.table) #if doing sctransform with newer version: library() |
Load Data and Create Seurat Object
Load data and create Seurat object
pbmc_data <- Read10X(data.dir = "filtered_feature_bc_matrix") pbmc <- CreateSeuratObject(pbmc_data) |
Check what the counts matrix looks like
Look at counts matrix
pbmc_data[c("CD3D", "TCL1A", "MS4A1"), 1:30]
##output
## CD3D 4 . 10 . . 1 2 3 1 . . 2 7 1 . . 1 3 . 2 3 . . . . . 3 4 1 5
## TCL1A . . . . . . . . 1 . . . . . . . . . . . . 1 . . . . . . . .
## MS4A1 . 6 . . . . . . 1 1 1 . . . . . . . . . 36 1 2 . . 2 . . . . |
Check what the seurat object looks like
Look at seurat object
pbmc An object of class seurat in project SeuratProject 33538 genes across 5155 samples. #right now, it just has counts, but as we process the data, many other slots will get added. |
Normalize Data Using SCTransform and Regress out Genes Related to Cell Cycle
Normalize data
s.genes <- cc.genes$s.genes g2m.genes <- cc.genes$g2m.genes pbmc <- CellCycleScoring(pbmc, s.features = s.genes, g2m.features = g2m.genes, set.ident = TRUE) pbmc$CC.Difference <- pbmc$S.Score - pbmc$G2M.Score pbmc <- SCTransform(pbmc, vars.to.regress = "CC.Difference") |
Identify Highly Variable Features
Identify Highly Variable Features and Generate Plots
pbmc <- FindVariableFeatures(pbmc, selection.method = "vst", nfeatures = 2000)
top10 <- head(VariableFeatures(pbmc), 10)
pdf("variableFeaturesPlot.pdf")
plot1 <- VariableFeaturePlot(pbmc)
plot2 <- LabelPoints(plot = plot1, points = top10, repel = TRUE)
CombinePlots(plots = list(plot1, plot2))
dev.off() |
Scale the Data
Scale data
all.genes <- rownames(pbmc) pbmc <- ScaleData(pbmc, features = all.genes) |
Select Dimensions using PCA (using variable features)
Perform PCA for Dimensionality Reduction
#Perform PCA for Dimensionality Reduction
pbmc <- RunPCA(pbmc, features = VariableFeatures(object = pbmc))
#Print and Examine PCA Results in Multiple Formats
pdf("findDimensions.pdf")
print(pbmc[["pca"]], dims = 1:5, nfeatures = 5)
VizDimLoadings(pbmc, dims = 1:2, reduction = "pca")
DimPlot(pbmc, reduction = "pca")
#Generate Principal Component Heatmaps through PC_30
DimHeatmap(pbmc, dims = 1, cells = 500, balanced = TRUE)
DimHeatmap(pbmc, dims = 1:15, cells = 500, balanced = TRUE)
DimHeatmap(pbmc, dims = 16:30, cells = 500, balanced = TRUE)
#Create Elbow Plot of Principal Components through PC_30
ElbowPlot(pbmc, ndims = 30, reduction = "pca")
dev.off() |
Cluster the cells (using the selected number of dimensions)
Cluster the Cells for the selected numeber of Principal Components
#Cluster the Cells for the selected numeber of Principal Components (we selected 14 for 5k Data) pbmc <- FindNeighbors(pbmc, dims = 1:14) pbmc <- FindClusters(pbmc, resolution = 0.5) #View the Cluster IDs of the First 5 Cells head(Idents(pbmc), 5) |
Visualize the clusters using tSNE or UMAP
Visualize using tSNE
pdf("tsne.pdf")
pbmcTSNE <- RunTSNE(pbmc, dims = 1:14)
DimPlot(pbmcTSNE, reduction = "tsne")
dev.off() |
Find markers for each cluster
Find Markers for Every Cluster and Print Top 2 Per Cluster
pbmc.markers <- FindAllMarkers(pbmcTSNE, only.pos = TRUE, min.pct = 0.25, logfc.threshold = 0.25) pbmc.markers %>% group_by(cluster) %>% top_n(n = 2, wt = avg_logFC) write.csv(pbmc.markers, file = "markersAll.csv") #Generate Heatmap of Top 10 Markers per Cluster top10 <- pbmc.markers %>% group_by(cluster) %>% top_n(n = 10, wt = avg_logFC) DoHeatmap(pbmcTSNE, features = top10$gene, size = 3) + theme(axis.text.y = element_text(size = 5)) + NoLegend() |
Identify cells expressing genes of interest
Identify Clusters for Marker Genes of Interest
markerGenes <- c("CD27", "CCR7", "CD8A", "CD8B", "IL7R", "GZMK", "CD79A", "CD37", "CD160", "NKG7", "GNL$
seuratClusters <- function(output, genes) {
for (gene in genes) {
print(output[output$gene == gene,]) }
}
seuratClusters(pbmc.markers, markerGenes)
#Visualize Feature Expression on TSNE Plot of Top Gene per Cluster
FeaturePlot(pbmcTSNE, features = c("LTB","S100A8","CCL5","NOSIP"))
FeaturePlot(pbmcTSNE, features = c("LINC02446","IGKC","GNLY","MT-CO3"))
FeaturePlot(pbmcTSNE, features = c("CST3","NKG7","KLRB1","LST1"))
FeaturePlot(pbmcTSNE, features = c("PPBP","AC084033.3"))
#Visualize Expression Probability Distribution Across Clusters for Top Gene per Cluster
VlnPlot(pbmcTSNE, features = c("LTB","S100A8","CCL5","NOSIP"))
VlnPlot(pbmcTSNE, features = c("LINC02446","IGKC","GNLY","MT-CO3"))
VlnPlot(pbmcTSNE, features = c("CST3","NKG7","KLRB1","LST1"))
VlnPlot(pbmcTSNE, features = c("PPBP","AC084033.3")) |
Label clusters as cell types based on expression of known marker genes.
Generate Labeled TSNE Plot
new.cluster.ids <- c("CD4+ Memory Cells", "CD16+ and CD14+ Monocytes", "Regulatory T Cells", "CD8+ T Cells", "N/A (Cluster 4)", "NK Cells", "B Cells", "Monocyte Derived Dendritic Cells", "N/A (Cluster 8)", "N/A (Cluster 9)", "Monocyte Derived Dendritic Cells", "N/A (Cluster 11)", "N/A (Cluster 12)", "Megakaryocyte Progenitors")
names(new.cluster.ids) <- levels(pbmcTSNE)
pbmcTSNE <- RenameIdents(pbmcTSNE, new.cluster.ids)
DimPlot(pbmcTSNE, reduction = "tsne", label = TRUE) |
Let's look at the results:
Download this zip file and unzip it on your computer to view the files (Most computers will automatically unzip files).
What do you do if you have multiple samples
If you have multiple samples (multiple cellranger output directories), read each one in and create individual seurat objects. Then, merged them into one object. You may want to use the project parameter to indicate condition information. You can also add metadata to do this.
From Seurat documention When merging, to easily tell which original object any particular cell came from, you can set the add.cell.ids parameter with an c(x, y) vector, which will prepend the given identifier to the beginning of each cell name. The original project ID will remain stored in object meta data under orig.ident.
data10<-Read10X(data.dir = "../../10/outs/filtered_feature_bc_matrix")
pbmc10<-CreateSeuratObject(counts = data10, project="dep")
data11<-Read10X(data.dir = "../../11/outs/filtered_feature_bc_matrix")
pbmc11<-CreateSeuratObject(counts = data11, project="control")
data14<-Read10X(data.dir = "../../14/outs/filtered_feature_bc_matrix")
pbmc14<-CreateSeuratObject(counts = data14, project="dep")
data16<-Read10X(data.dir = "../../16/outs/filtered_feature_bc_matrix")
pbmc16<-CreateSeuratObject(counts = data16, project="control")
pbmc.combined<-merge(pbmc10, y=c(pbmc11,pbmc14,pbmc16),add.cell.ids = c("10","11","14","16), project = "pbmc", merge.data = TRUE) |
Single Cell Proportion Test
In order to identify differences in proportions of cell types between conditions, scProportionTest can be run. It reads in a seurat object and runs a permutation test to calculate a p-value for each cluster and a magnitude difference between conditions. It also generates a plot showing the pvalues and difference. It can be used to identify enriched or depleted cell types between conditions.
library("scProportionTest")
#read in seurat object, saved as a RData file
load("seurat.RData")
#read in seurat object (pbmc), assuming that pbmc has multiple samples with different condition identities
prop_test <- sc_utils(pbmc) |
#run permutation test to compare between cells with orig.ident as 'Con' and ident 'Treatment' #run it for every cluster prop_test <- permutation_test( prop_test, cluster_identity = "custom_clusters", sample_1 = "Con", sample_2 = "Treatment", sample_identity = "orig.ident" ) |
#Generate plot permutation_plot(prop_test) |