Setting colour palettes correctly for correlation plots

Question

[exin]

Newbie R question:

I currently use the standard blue-red palette for plotting a correlation plot:

> col<- colorRampPalette(c("blue", "white", "red"))(20)

but my data only covers a R value or ~0.2 to 1, hence white is set at ~0.4, which is misleading.

How do I select only "some blue and all red" of the spectrum, still with 20 divisions?

Undesired scale colours & associated plot: https://ibb.co/Hx6bzdM

[Friederike]

can you provide the entire code you use for the heatmap generation, not just the color part? (different heatmap functions may have different capabilities of tuning the colors)

[exin]

Here we go:

cts<- read.table("cor_gene.txt",header=T,row.names=1,sep="\t")
cts<- cts[rowSums(cts)>0,]
cts$var<- apply(cts, 1, var)
ctsv<- cts[cts$var >= quantile(cts$var, c(.95)), ]
ctsv$var <- NULL
resv<-cor(ctsv,method="pearson")

library("RColorBrewer")
library("pheatmap")

col<- colorRampPalette(c("blue", "white", "red"))(20)
pheatmap(resv, col=col, cluster_cols=T, cluster_rows=T, 
         show_rownames=T, scale = "none", border_color=F,
         fontsize_row=5.2, fontsize_col=5.2, fontsize=8, 
         cellheight=5, cellwidth=5, clustering_method = "complete")

Answer 1

[Jean-Karim Heriche]

Check the breaks option. Have a look at this pheatmap tutorial.

[exin]

Looks like what I need... Thank you! I'll look into that.