Resizing

When you saw your heatmap pop up, you may (or may not) have noticed that some of the labels were truncated. That’s because the entire plot is a bit too large for the viewing area. So we need to shrink it a bit. To do this we will modify the margins argument. You will pass a vector of 2 numbers (example: c(5,5)) that will add space below and to the right of the heatmap. Play around with these two numbers and see how it impacts your heatmap.

I’ve also started to add comments to each line of the heatmap.2 code. Anything after the # is a comment and is not interpreted by the machine, they are simply used to make the code more reader-friendly:

heatmap.2(as.matrix(micro_data), # data frame a matrix
          margins = c(6,15)) # Adds margins below and to the right

Remove the density/trace lines

You will also notice some light blue (cyan) lines that are drawn through the heatmap and the legend. These can be useful (especially in the legend), becaues it give you an idea of how frequently values appear in your data. For example, in the above figure legend, you can see that there are a lot of values that are 0. Again, useful, but not really needed…so let’s remove them:

heatmap.2(as.matrix(micro_data), # data frame a matrix
          margins = c(6,15), # Adds margins below and to the right
          density.info = "none", # Remove density legend lines
          trace = "none") # Remove the blue trace lines from heatmap

Remove Row Dendrogram

While it might be important to see how the varous bacteria cluster, most microbiome researchers really care about how samples cluster. So, let’s remove the clustering from the rows, and only keep the column clustering and dendrogram.

heatmap.2(as.matrix(micro_data), # data frame a matrix
          margins = c(6,15), # Adds margins below and to the right
          density.info = "none", # Remove density legend lines
          trace = "none", # Remove the blue trace lines from heatmap
          Rowv = FALSE, # Do not reorder the rows
          dendrogram = "col") # Only plot column dendrogram

Add gridlines

As long as you don’t have a ton of samples/bacteria, adding gridlines to your heatmap can really spruce it up! With the heatmap.2 function, there are two arguments called colsep and rowsep. You want to pass a vector of numbers that tell R where to add vertical and horizontal separating lines:

heatmap.2(as.matrix(micro_data), # data frame a matrix
          margins = c(6,15), # Adds margins below and to the right
          density.info = "none", # Remove density legend lines
          trace = "none", # Remove the blue trace lines from heatmap
          Rowv = FALSE, # Do not reorder the rows
          dendrogram = "col", # Only plot column dendrogram
          colsep=1:nrow(micro_data), # Add vertical grid lines
          rowsep=1:nrow(micro_data), # Add horizontal grid lines
          sepcolor = "black") # Color gridlines black

Adding a custom color scale

If I took the above heatmap image to my boss and said “Here…heatmap!”, he would probably look at it and say, “Ryan…this is not very informative!” And he’d be right, because the majority of our values fall in 0-20 range..and if you look at your legend, there is not much color change in that range…it’s all red. So we need to make a custom color range. This is a bit more advanced, but stay with it and I promise your heatmaps will look amazing AND informative!

First we need to create a custom color gradient. I really like the viridis color gradient, as it’s not only pretty but it’s easy to read for those with colorblindness. Let’s install and load the viridis package:

install.packages("viridis")
library(viridis)

Now let’s feed the viridis::viridis_pal() function into the col argument:

heatmap.2(as.matrix(micro_data), # data frame a matrix
          margins = c(6,15), # Adds margins below and to the right
          density.info = "none", # Remove density legend lines
          trace = "none", # Remove the blue trace lines from heatmap
          Rowv = FALSE, # Do not reorder the rows
          dendrogram = "col", # Only plot column dendrogram
          colsep=1:nrow(micro_data), # Add vertical grid lines
          rowsep=1:nrow(micro_data), # Add horizontal grid lines
          sepcolor = "black", # Color gridlines black
          col = viridis::viridis_pal()) # Make colors viridis

That definitely looks better, but we should still add a little more color separation in the 0-20 range. To do this, we will give the breaks argument a vector. Add more breaks between 0-20 so that more colors are used within this range of values. Again, feel free to play around with this vector of numbers and see how it changes the plot. Here is my custom scale (using the built in seq() function):

c(0:20, seq(from = 25, to = max(micro_data) + 10, by = 10))

##  [1]  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 25 35
## [24] 45 55 65 75 85

The seq function creates a vector from 25 to the max value in my data set (84) + 10 in intervals of 10. I added the + 10 just to give a little bit of buffer at the end of my vector to ensure the max value was encompassed. Now feed that vector into the col argument:

heatmap.2(as.matrix(micro_data), # data frame a matrix
          margins = c(6,15), # Adds margins below and to the right
          density.info = "none", # Remove density legend lines
          trace = "none", # Remove the blue trace lines from heatmap
          Rowv = FALSE, # Do not reorder the rows
          dendrogram = "col", # Only plot column dendrogram
          colsep=1:nrow(micro_data), # Add vertical grid lines
          rowsep=1:nrow(micro_data), # Add horizontal grid lines
          sepcolor = "black", # Color gridlines black
          col = viridis::viridis_pal(), # Make colors viridis
          breaks = c(0:20, seq(25, max(micro_data) + 10, 10))) # Custom scale of breaks

Save you Heatmap

There are many “devices” that can be used in R to save your plot, but I will go with the png() function. To save a heatmap, you first need to “turn on” the device, plot your heatmap, then you turn the device off. For example:

# Turn on the png device
png(filename = "yourplot.png", width = 5, height = 7, units = "in", res = 300)

# Plot your heatmap
heatmap.2(your_heatmap)

# Turn off png device
dev.off()

The filename argument will be the name of the plot when it is saved to your computer. If you simply give it a name (like I did above), the plot will be saved to whatever your current working directory is. What is my current working directory? Good question. Type this into your R console and see what the output is:

getwd()

Maybe your output looks something like this:

"/Users/yourname/Desktop"

This means that once you save your plot, it will be saved to your desktop. So if you ever loose your plots, just run this command and then look in that directory for your plot. Ok, let’s save our heatmap:

png(filename = "myheatmap.png", width = 7, height = 5, units = "in", res = 300)

heatmap.2(as.matrix(micro_data), # data frame a matrix
          margins = c(6,15), # Adds margins below and to the right
          density.info = "none", # Remove density legend lines
          trace = "none", # Remove the blue trace lines from heatmap
          Rowv = FALSE, # Do not reorder the rows
          dendrogram = "col", # Only plot column dendrogram
          colsep=1:nrow(micro_data), # Add vertical grid lines
          rowsep=1:nrow(micro_data), # Add horizontal grid lines
          sepcolor = "black", # Color gridlines black
          col = viridis::viridis_pal(), # Make colors viridis
          breaks = c(0:20, seq(25, max(micro_data) + 10, 10))) # Custom scale of breaks

dev.off()

The units arguments means to save a 7x5 inch image with a pixel resolution of 300 pixels per inch (ppi). Hopefully you can go into your working directory and open up your image. If so…send it to your advisor and be prepared for a “Wow, this is great!” email in response. I hope this helps and please let me know if you have any questions!