Plot heatmap using row_splitting but splitting should based on a column in dataset - ComplexHeatmap

Question

[TJay]

I have a dataset with following format.

  X        M1.    M2.     ...    new_cell_type
cac1.       1.5        1      ...     T
cac2.       1.4      1.4    ...      M
cac3.       0.5      1.1    ...      T
cac4.       0.1      1      ...      N

I want to plot a heatmap based on the above data where rows of the heatmap as cells with cell grouping(with cell_type) and columns as M1,M2,M3... I tried to split the cells with cell type pass that as a matrix to plot heatmap using ComplexHeatmap. But it doesn't work. Any suggestions for this?

data_c <- read.csv('./output/a.csv')
cell_type <- read.csv(file = 'b.csv')
merged_data <- merge(data_c, cell_type,by = "X")

df_list <- split(merged_data, merged_data$new_cell_type)
df_list <- as.matrix(df_list)

heat_col <- colorRamp2(c(0, 0.5, 1), c("blue", "white", "red"))

hm <- Heatmap( df_list, 
    col = heat_col, 
    row_split = merged_data$new_cell_type, 
    show_row_names = TRUE,
    cluster_rows = TRUE, 
    cluster_columns = TRUE,
   )
  draw(hm)

I'm expecting this kind of heatmap where rows should be cells and left bar is representing celltype. columns should be M1,M2,M3, etc.

[rpolicastro]

Instead of splitting your dataset you want to make an annotation object via rowAnnotation and then pass it to the left_annotation argument.

[TJay]

Any guidance to make annotation object?

[rpolicastro]

This is all detailed in the vignettes for ComplexHeatmap. You should read through them first and come back with any specific questions you may still have afterwards.

[TJay]

Yes. I went through that and changed my code as following. Still it's not giving the correct output.

UPDATE.

 data_c <- read.csv('./output/a.csv')
 cell_type <- read.csv(file = 'b.csv')

 merged_data <- merge(data_c, cell_type,by = "X")
 mat <- data.matrix(merged_data[, 2:(ncol(df)-1)])
 rownames(mat) <- merged_data$X
 cell_types <- factor(merged_data$new_cell_type, levels=c("B", "DC", "E","Myel", "NK", "Normoblast","Prog", "T"))
 df_cell_types <- data.frame(cell_type = cell_types)

row_anno <- HeatmapAnnotation(
#df = data.frame(cell_type = cell_types),
df = df_cell_types,  
#width = unit(1, "cm"),
#height = unit(0.5, "cm"),
col = list(cell_type = c(B = "#E41A1C", DC = "#377EB8", E = "#4DAF4A", Myel = "#E41A1C", NK = "#377EB8", Normoblast = "#4DAF4A",Prog ="#E41A1C", T= "#377EB8")),
#show_legend = TRUE
which = "row"
)

hm <- Heatmap(mat, 
name = "Expression", 
col = colorRamp2(c(-1, 0, 1), c("blue", "white", "red")),
show_row_names = TRUE,
#clustering_distance_rows = "euclidean",
#clustering_distance_columns = "euclidean",
#row_dend_reorder = TRUE,
row_title = "Cells",
column_title = "Modules",
right_annotation = row_anno
)

draw(hm)

[Trivas]

First, set show_row_names = FALSE. You might want to play around with the col = col_fun portion of the vignette to change your scale to be something that more closely represents your data.