Bio-Themed Visuals

Collection of major bio-themed data visuals   by Amal Katrib

(1) Violin Plot

A box plot and kernel density plot hybrid that shows summary statistics as well as the full distribution of the data

Below is a sample code that can be used to generate violin plots in R:

# load the appropriate packages
library(ggplot2)

# generate plots
plist = list()
x = 1
for (i in unique(data$geneName)) {
  p = ggplot(data[data$geneName %in% i, ], aes(x = group1, y = value)) +
        geom_violin(scale = "count", position = position_dodge(width = 1), trim = F) +
        geom_boxplot(aes(x = group1, y = value), notch = F) +
        geom_point(position = position_jitterdodge(jitter.width = 0.5), aes(color = group2)) +
        geom_vline(xintercept = c(x,y)) +
        labs(x = "", y = ""))

  plot_list[[x]] = p
  x = x + 1  }
names(plist) = unique(data$geneName)

# save plots
lapply(1:length(plot_list), function(i) {
           png("violionPlot.png"), 5, 5, res = 300, units = "in")
           print(plot_list[[i]])
           dev.off() })

---

(2) Heatmap

A hierarchical clustering visual with a color scale-rendition of numerical data to help reveal underlying patterns

I recommend using the heatmap.3() function in R so you can include multiple row and column side bars with added sample and gene info. Data inputs, and their corresponding formats, include:

• "data" matrix   log-/variance stabilization-transformed normalized read counts (when used in next-gen seq)
• "clab" matrix   color mapping of sample of info matrix

	sample 1	sample 2	sample 3	sample 4
gene 1	3	10	9	5
gene 2	9	4	6	10
gene 3	3	6	6	9
gene 4	8	6	8	10

	infoColor 1	infoColor 2	infoColor 3	infoColor 4
sample 1	red	yellow	orange	darkblue
sample 2	red	green	black	darkred
sample 3	blue	yellow	orange	darkblue
sample 4	blue	yellow	black	darkblue

Below is a sample code that can be used to generate heatmaps in R:

hr <- hclust(as.dist(1-cor(t(data), method="pearson")), method="average")
hc <- hclust(as.dist(1-cor(data, method="pearson")), method="average")

heatmap.3(data,
                  Rowv = as.dendrogram(hr), Colv = as.dendrogram(hc),
                  dendrogram = "both", col = palette, ColSideColors = clab, key = TRUE)

# select a data-representative color palette
palette <- colorRampPalette(c("yellow3","white","darkblue"))

Depending on what you intend to visualize, data can be scaled to mean = 0 & standard deviation = 1 either by:

• Setting the scale parameter in the heatmap function using heatmap.3(scale = "row" )
• Directly scaling the matrix content using t(scale(t(data)))  

• Re-arrange columns in the heatmap to best convey your message, either by:
  • Maintaining the original sample order
  • Using unsupervised hierchical clustering of samples
• Pay attention to the color scheme:
  • Use Diverging Palettes such as red-blue or yellow-blue if you want to have 2 contrasting colors that represent variation from a reference value. This is often used in heatmaps when representing differential analysis results
  • Use Sequential Palettes such as white-lightgrey-darkgrey-black if you want to represent sequential (increasing / decreasing) data such as age and height
  • Use Categorical Palettes such as red-black-yellow-orange if we want to represent categorical data such as gender and disease state
  • Select a color scheme that color-blind individuals can readily see.AVOID RED-GREEN
  • Avoid excessive inclusion of colors so as to not confuse your audience
  • Consider the well-perceived "viridis" color scale: install.packages("viridis")