Hey folks,

We had a lot of fun last week with my first workshop on the theory of data visualization! If this is something that you’d be interested in participating in let me know. At this point, I don’t have anything scheduled. So, if you have suggestions for days or times, please let me know

This week I have a fun figure to share with you from a paper recently published in Nature Microbiology, titled, “Candida auris skin tropism and antifungal resistance are mediated by carbonic anhydrase Nce103”. Figure 2a caught my eye

If you’re a microbiologist, you might notice that it resembles a microtitre plate that are often used for performing antibiotic susceptibility testing in solution. For the uninitiated, imagine you have a 96-well plate. Each column has a different step in a dilution series of antibiotic added to growth media. In this case there are 2-fold dilutions of Amphotericin B applied across the columns. All columns of a given row are inoculated with a different strain of bacteria. After incubating the plate you can score the dilution where bacteria stop growing. This is the “minimum inhibitory concentration” or “MIC”. You can see an example of this as performed on agar in panel e of this figure. The method used by these authors allowed them to score the amount of growth relative to the column with no antibiotic added. As I mentioned, if your eyes look like mine, you can see this looks like a microtitre plate. How would we make this in R?

Well, if your eyes look like mine, this panel resembles a heatmap. Instead of having rectangular tiles, this panel has circles that are filled according to the relative growth level. As an added wrinkle, this experiment incubated the assay at three different CO₂ levels making for three facets across the panel.

Thankfully, the authors made their data available as a Microsoft XLSX workbook. If you open tab “F2A”, you’ll see the data for this figure. The data frame is already “tidy” with columns for the strain, amount of antibiotic, CO₂ concentration, and relative growth. As we’ve seen in recent videos, we can read these workbook pages in to R using the read_excel() function from the {readxl} package.

From here I think it is relatively straightforward to create the basic figure. On the x-axis we map the concentration of the antibiotic. On the y-axis we map the strain. We would then map the relative growth to the fill color. We can use geom_point() to draw the plot. I’d use my favorite plotting symbol^TM, 21, which will allow us to have one color for the fill and one for the border. A more challenging step is denoting the MIC for each strain. This is denoted by a red border to the circle and the “MIC” text in the middle of the circle. I’d map whether the concentration was the MIC to the color of the circle with the MIC getting a red color and everything else getting black. To insert the “MIC” in the circle, I’d use geom_text(). Not too bad, eh?

I feel like another unique element of the figure are the text elements. First, consider the legend. The title is rotated. I think this can be done using the theme() function with the legend.title argument. This argument will take the element_text() function, which has an angle argument that we can set to 90. Next, the x-axis text has the dilutions at a 45 degree angle. Here again, we can set the angle argument for element_text() when assigned to the axis.text.x argument of theme(). We will likely have to fiddle with the hjust and vjust arguments in element_text() to get things to look right. The y-axis has a mix of italicization and Greek letters in the text. We will likely have to set the order of the strains using factor() and can also change the labels of the strains to incorporate unicode and markdown to add the delta characters and italicization where needed. Finally, the facet strips and x-axis title will also need some markdown to add sub and super script characters. All of this markdown can be rendered using the {ggtext} package’s element_markdown() function in place of the element_text() function.

Finally, I’d be chickening out if I didn’t mention the vertical lines between the three facets. I feel like we’ve done something like this in the past using annotate() with geom = "vline" and the layout argument that new to {ggplot2} v.4.0. I think this should work 🤓

What do you think? Can you pull this off on your own? Give it a try! I’ll be recreating it along with some tweaks to make it better (IMHO) next Wednesday during a livestream on YouTube. Also, stay tuned for Monday when I’ll release a critique of this plot discussing what I like or don’t like.

​

​