Plotting dimension reductions

Dimension reductions can be plotted by function plot_scdata():

plot_scdata(scRNA_int, pal_setup = pal)
UMAP plotting, colored by clusters

UMAP plotting, colored by clusters

There are three optional parameters for plot_scdata(): color_by, split_by, and pal_setup. As for the color_by parameter, the function will color different "seurat_clusters" by default, and it can be changed to any factors in the metadata, like "sample" or "group":

plot_scdata(scRNA_int, color_by = "group", pal_setup = pal)
UMAP plotting, colored by groups

UMAP plotting, colored by groups

If split_by parameter is specified as a factor in the metadata, the plotting will be split by that factor:

plot_scdata(scRNA_int, split_by = "sample", pal_setup = pal)
UMAP plotting, split by samples

UMAP plotting, split by samples

Similar to the plot_qc() function, the pal_setup parameter can be RColorBrewer palette names, palette setup dataframe, or manually specified color vector.

plot_scdata(scRNA_int, pal_setup = "Dark2")
UMAP plotting, colored by clusters, RColorBrewer Dark2 palette

UMAP plotting, colored by clusters, RColorBrewer Dark2 palette 

plot_scdata(scRNA_int, color_by = "sample", pal_setup = c("red","orange","yellow","green","blue","purple"))
UMAP plotting, colored by clusters, mannually specified palette

UMAP plotting, colored by clusters, mannually specified palette

Plotting statistics

The count and proportion statistics of clustering can be plotted by function plot_stat(), the plot_type parameter must be provided as one of the three values: "group_count", "prop_fill", and "prop_multi". Their plots are shown below:

plot_stat(scRNA_int, plot_type = "group_count")

plot_stat(scRNA_int, "group_count", group_by = "seurat_clusters", pal_setup = pal)

plot_stat(scRNA_int, plot_type = "prop_fill", 
          pal_setup = c("grey90","grey80","grey70","grey60","grey50","grey40","grey30","grey20"))

plot_stat(scRNA_int, plot_type = "prop_multi", pal_setup = "Set3")

The group_by parameter uses "sample" as the default grouping variable, and it can be specified as other factors in the metadata (e.g. "group").

plot_stat(scRNA_int, plot_type = "prop_fill", group_by = "group")

plot_stat(scRNA_int, plot_type = "prop_multi", group_by = "group", pal_setup = c("sienna","bisque3"))

Plotting heatmap

The plotting of heatmap requires cluster markers to be found by Seurat:

markers <- FindAllMarkers(scRNA_int, logfc.threshold = 0.1, min.pct = 0, only.pos = T)

Then, the top genes in each cluster are plotted by plot_heatmap(). The default value of number of genes plotted in each cluster n is 8. In the heatmap, each row represents a gene and each column a cell. The cells can be sorted by sort_var can it is set to c("seurat_clusters") by default, meaning the cells are sorted by cluster identity. Multiple variables can be specified in sort_var and the cells will be sorted by the order of the variables. The bars above the heatmap are annotation bars and can show categorical or continuous variables in the metadata by specifying the anno_var parameter, with variable names as a character vector. The anno_colors parameter is a list that specifies the annotation colors for corresponding annotation variables hence it should be the same length as anno_var. It is recommended that proper color palettes are used for categorical and continuous variables. As before, RColorBrewer palettes and manually specified palettes are supported, and a three-color vector can be used for continuous variable annotation.

plot_heatmap(dataset = scRNA_int, 
              markers = markers,
              sort_var = c("seurat_clusters","sample"),
              anno_var = c("seurat_clusters","sample","percent.mt","S.Score","G2M.Score"),
              anno_colors = list("Set2",                                             # RColorBrewer palette
                                 c("red","orange","yellow","purple","blue","green"), # color vector
                                 "Reds",
                                 c("blue","white","red"),                            # Three-color gradient
                                 "Greens"))

Furthermore, hm_limit and hm_colors are used to specify the color gradient and limits of the main heatmap tiles.

plot_heatmap(dataset = scRNA_int,
             n = 6,
             markers = markers,
             sort_var = c("seurat_clusters","sample"),
             anno_var = c("seurat_clusters","sample","percent.mt"),
             anno_colors = list("Set2",
                                c("red","orange","yellow","purple","blue","green"),
                                "Reds"),
             hm_limit = c(-1,0,1),
             hm_colors = c("purple","black","yellow"))

GO Analysis

The GO analysis results can be plotted by plot_cluster_go() and plot_all_cluster_go(). The former plotted one specific cluster while the latter iterates all clusters. The topn parameter in plot_cluster_go() specifies the number of top genes for GO analysis and the default value is 100. The org parameter specifies the organism, and "human" and "mouse" are the accepted values. plot_all_cluster_go() is the wrapper for plot_cluster_go() and the latter is again a wrapper for clusterProfilter::enrichGO(). Hence, the ... parameters can be passed into inner functions.

plot_cluster_go(markers, cluster_name = "1", org = "human", ont = "CC")

plot_all_cluster_go(markers, org = "human", ont = "CC")

Plotting Measures

The measures are defined as continuous variables in the metadata as well as gene expression values. The plot_measure() and plot_measure_dim() summarize these variables as box/violin plots and dimension reduction plots, respectively. Parameters like group_by, split_by, and pal_setup can be used similarly as described above.

plot_measure(dataset = scRNA_int, 
             measures = c("KRT14","percent.mt"), 
             group_by = "seurat_clusters", 
             pal_setup = pal)

plot_measure_dim(dataset = scRNA_int, 
                 measures = c("nFeature_RNA","nCount_RNA","percent.mt","KRT14"))

plot_measure_dim(dataset = scRNA_int, 
                 measures = c("nFeature_RNA","nCount_RNA","percent.mt","KRT14"),
                 split_by = "sample")

GSEA Analysis

To perform GSEA analysis, we will first find the differentially expressed genes (DEGs) and related measures by find_diff_genes(). Then, the ranked list will be input for GSEA analysis by test_GSEA(). (Note: It may take Seurat a long time to find DEGs. Parallel processing by package future is recommended.). Finally, the output can be plotted by plot_GSEA(), with additional parameters provided for adjusted p-value cutoff and color gradients.

de <- find_diff_genes(dataset = scRNA_int, 
                      clusters = as.character(0:7),
                      comparison = c("group", "CTCL", "Normal"),
                      logfc.threshold = 0,   # threshold of 0 is used for GSEA
                      min.cells.group = 1)   # To include clusters with only 1 cell

gsea_res <- test_GSEA(de, 
                      pathway = pathways.hallmark)
plot_GSEA(gsea_res, p_cutoff = 0.1, colors = c("#0570b0", "grey", "#d7301f"))