R: Row scaling not working correctly for heatmap

Question

I am currently creating a heatmap for some data where the values for each item (OTU) vary greatly, so I am using the scale="row" option.

I assumed that this would utilise the same colour gradient for each row so the maximum colour would always be present (dark green), as would the minimum (white). However as you can see below some go from white, others from a light green, and go to a dark or darkish green.

OTU table

Here is the code I have been using:

library(gplots)
library(RColorBrewer)
data <-read.csv("Heatmap_data.csv", comment.char="#")
rnames <- data[,1]                            # assign labels in column 1 to "rnames"
mat_data <- data.matrix(data[,2:ncol(data)])  # transform column 2-5 into a matrix
rownames(mat_data) <- rnames                  # assign row names


my_palette <- colorRampPalette(c("white", "dark green"))(n = 1090)


heatmap(mat_data, Rowv=NA, Colv=NA, col = my_palette, scale="row", margins=c(23,15))

Many of the values which are off-white are actually 0s as you can see in the data here:

MB.11   MB.8    MB.7    MB.12   MB.9    MB.10   MB.4    MB.6    MB.5    MB.3    MB.2    MB.1    K2295   K229456 K2299   K2297   K2298   K2293   K2294   K2292   K2290   K2291
OTU_43  23.5019505852   4.3947267864    6   18.3651985319   3.0748596382    5.5431848853    71.853471596    24.1857966978   12.6422018349   13.2893694289   46.5896990741   21.0355475222   38.5817456801   7.723156853 55.9703296703   6.1243646222    72.4165876777   74.0147572816   46.6683375104   87.3876576884   23.2358576642   33.7528164347
OTU_48  27.7750325098   36.4134505159   9   55.0955955956   9.7370555209    8.8690958165    108.8913435528  25.8199721504   23.8256880734   10.3060415979   42.7864583333   4.7020635638    3.6397873283    3.2518555171    69.8124542125   0   3.1146919431    26.7984466019   124.2201336675  355.7125809751  38.2267335766   63.8762977689
OTU_162 8.0120286086    11.3007260222   2   8.7714381048    2.5623830318    3.3259109312    35.5563570784   19.6101054307   12.6422018349   16.0014856388   57.0486111111   31.9245368278   4.367744794 3.2518555171    6.3192307692    3.8973229414    12.4587677725   29.3506796117   11.6670843776   35.8513467439   47.2212591241   45.3666887564
OTU_64  16.0240572172   0   36  35.6339673006   0.5124766064    5.5431848853    3.7037871957    0   1.4587155963    0   24.7210648148   0   1.4559149313    0   0   0   0   0   0   3.3610637572    0   2.1776010603
OTU_60  0   61.5261750096   23  0   42.0230817218   0   0   0   8.752293578 7.5939253879    3.8032407407    11.6314203946   21.8387239699   13.8203859476   8.1247252747    0   22.5815165877   0   3.4314954052    14.0044323219   8.2449817518    3.6293351005
OTU_116 136.7386215865  8.7894535728    15  30.9741408075   23.0614472863   103.6575573549  0   0   0   0   0   0   31.3021710235   8.1296387927    28.586996337    28.9515418502   49.83507109 0   0   0   0   0
OTU_86  7.477893368 26.3683607184   1   2.1928595262    21.5240174672   16.6295546559   2.9630297565    0.653670181 0   0   13.3113425926   0   0   0   0.3009157509    0   0   0   0   2.8008864644    0.7495437956    1.0888005302
OTU_534 134.0679453836  0   15  0   26.6487835309   180.1535087719  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0
OTU_75  81.7226918075   0   6   0   12.2994385527   121.3957489879  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0
OTU_176 26.7067620286   0   0   19.461628295    26.6487835309   9.423414305 0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   7.9845372211

Any help would be greatly appreciated!

All values in the scaled matrix are mapped on the color scheme you defined not row by row. Also I recommend to use diverging color scheme for scaled data. — Lerong
@Lerong The scalled matrix being mapped onto the color scheme is what I want, however during the scalling, samples with a value of 0 are being scalled to be off-white for some reason. An example of this is in the top row the 7th till the last column all have values of 0 (this row is OTU_176 in the data table) and yet are all plotted as off-white instead of white. — thh32

Jakub Buček Jakub Buček · Accepted Answer · 2018-04-11T15:02:20

You scale the color for each row separately. The

heatmap(log(mat_data+1), Rowv=NA, Colv=NA, col = my_palette, scale="none", margins=c(23,15))

should do the job.

The scale="none" suppresses any scaling and log(mat_data+1) shrinks the range to more suitable values.

R: Row scaling not working correctly for heatmap

1 Answers