Ordering row with high expression at the top rather than the bottom

Question

[synat.keam]

Hi All,

hope you are well. been trying to arrange order of row in decreasing order, but did not work. Basically, I like to have row with high expression at top and then arrange all row in decreasing order based on their relative expression. I worked previously, but did not work this time. Tried to solve it, but could not fix it. these are fake dataset. have a good day!

library("readxl")
# xls files
cibersort <- read_excel("Cibersort_timecourse_raw.xlsx")

## Convert each column to percentage ================================

## step 1 is to write a function ================

## source https://stackoverflow.com/questions/30457951/convert-columns-i-to-j-to-percentage

myfun <- function(x) {
  if(is.numeric(x)){ 
    ifelse(is.na(x), x, paste0(round(x*100L, 1))) 
  } else x 
}

## Apply mutate each ===========================================

Cibersort_percent<- cibersort %>% mutate_each(funs(myfun)) 

## Perform hierachical clustering =========================

dim(Cibersort_percent)
head(Cibersort_percent)

## Transpose =================================

Cibersort.Tran<- as.data.frame(t(Cibersort_percent[-1]))
names(Cibersort.Tran)= Cibersort_percent$Mixture
rownames(Cibersort.Tran)=names(Cibersort_percent[-1])

## load phenotype data =================

Pheno <- read_excel("Phenotype.xlsx")

## Selection relevant ones =================

phenotype<- Pheno[1:2]

## Scale the numeric variables ==================================

Cibersort.numeric<- Cibersort.Tran %>% mutate_if(is.character,as.numeric)

## Feature scaling ==============================

df.scale<- scale(Cibersort.numeric)

## math row/column of datframe and phenotypic data 

Group_df = data.frame("Group" = phenotype$Group)
rownames(Group_df) = colnames(Cibersort.Tran) # name matching

# plot pheatmap ===========================

library(pheatmap)

pheatmap(df.scale, cluster_cols = T, annotation = select(Group_df, Group))

[Basti]

It is difficult to answer without data, but you could convert Group_df$Group to factor and set the levels in the order you want without clustering. Something like Group_df$Group=factor(Group_df$Group,levels=c("M0.Macrophage","M1.Macrophage", ...)) and then remove the clustering of rows with cluster_rows=F : pheatmap(df.scale, cluster_cols = T, cluster_rows=F, annotation = select(Group_df, Group))

[synat.keam]

Dear Basti,

Thank you! I will experiment your code chunk. many thanks,

sk,