Exercise 7 — Advanced RNA-seq Visualization

In real RNA-seq analysis, base R plots are rarely used for final figures.

Instead, we use:

  • ggplot2 (for publication-quality plots)
  • pheatmap (for heatmaps with clustering)
  • patchwork / gridExtra (to combine plots)

This exercise will teach you how to create:

  • modern RNA-seq visualizations
  • structured multi-panel figures
  • publication-style plots

🎯 Learning Objectives

By the end of this exercise, you should be able to:

  • Use ggplot2 for gene expression plots
  • Create boxplots and barplots using ggplot2
  • Use pheatmap for clustered heatmaps
  • Combine multiple plots into one figure
  • Compare visualization styles (base R vs ggplot2)
  • Interpret multi-panel RNA-seq figures

📦 Setup

Task 0.1 — Load libraries

library(ggplot2)
library(pheatmap)
library(gridExtra)

🧬 Data Setup

We will reuse a simplified RNA-seq dataset.

Task 1.1 — Expression matrix

genes <- c("TP53","EGFR","MYC","BRCA1","ACTB","GAPDH","VEGFA","CDKN1A")

counts <- matrix(c(
  120,130,125,200,210,220,
  300,290,310,450,470,490,
  80,75,90,150,160,170,
  500,520,510,800,820,850,
  1000,980,1020,1100,1120,1150,
  900,920,910,950,970,960,
  150,160,155,300,320,310,
  200,190,210,400,420,410
), nrow = 8, byrow = TRUE)

rownames(counts) <- genes
colnames(counts) <- c("C1","C2","C3","T1","T2","T3")

Part 1 — Converting to Long Format (IMPORTANT)

ggplot2 works best with long-format data.

Task 1.2

Convert matrix → data frame:

df <- data.frame(
  Gene = rep(rownames(counts), times = ncol(counts)),
  Sample = rep(colnames(counts), each = nrow(counts)),
  Expression = as.vector(counts)
)

Task 1.3

Add condition column:

df$Condition <- ifelse(grepl("C", df$Sample),
                       "Control",
                       "Treatment")

Part 2 — ggplot2 Visualizations

Task 2.1 — Gene expression boxplot

ggplot(df, aes(x = Gene, y = Expression)) +
  geom_boxplot()

Task 2.2 — Improve aesthetics

Add:

  • fill by condition
  • axis rotation
  • title

Task 2.3 — Sample-level comparison

Create boxplot:

  • x = Sample
  • y = Expression
  • fill = Condition

Task 2.4 — Interpretation

Answer:

  • which group shows higher expression?
  • are distributions consistent across samples?

Part 3 — Barplots in ggplot2

Task 3.1 — Mean expression per gene

First compute:

gene_means <- rowMeans(counts)

Convert to data frame:

plot_df <- data.frame(
  Gene = names(gene_means),
  MeanExpression = gene_means
)

Task 3.2 — Barplot

ggplot(plot_df, aes(x = Gene, y = MeanExpression)) +
  geom_bar(stat = "identity")

Task 3.3 — Improve visualization

Add:

  • color fill
  • rotated labels
  • title

Part 4 — pheatmap Visualization

Task 4.1 — Basic heatmap

pheatmap(counts)

Task 4.2 — Scaled heatmap (IMPORTANT)

pheatmap(scale(counts))

Task 4.3 — Interpretation

Answer:

  • which genes cluster together?
  • do Control and Treatment separate?

Part 5 — Combining Plots (VERY IMPORTANT)

Task 5.1 — Create ggplot objects

p1 <- ggplot(plot_df, aes(x = Gene, y = MeanExpression)) +
  geom_bar(stat = "identity") +
  theme(axis.text.x = element_text(angle = 90))

p2 <- ggplot(df, aes(x = Gene, y = Expression)) +
  geom_boxplot()

Task 5.2 — Combine plots

Using grid.arrange:

grid.arrange(p1, p2, ncol = 2)

Task 5.3 — Interpretation

Answer:

  • what does each plot tell you?
  • why is combining plots useful?

Part 6 — Multi-panel Figure Thinking

Task 6.1

Create:

  • one plot showing gene means
  • one plot showing sample distributions
  • one heatmap

Task 6.2

Combine them into one figure

Task 6.3

Reflect:

  • how does this resemble a paper figure?

Part 7 — Biological Interpretation

# 1. Which genes are most variable?
# 2. Do Control and Treatment separate visually?
# 3. Why is scaling important in heatmaps?
# 4. Why do we use ggplot2 instead of base R?
# 5. What does this tell you about RNA-seq structure?

Bonus Challenge

B1

Create a ggplot heatmap-style visualization using geom_tile()

B2

Order genes by mean expression before plotting

B3

Make a publication-ready figure:

  • title
  • clean theme
  • rotated labels
  • consistent colors

🧠 Key Concepts Learned

  • ggplot2 grammar of graphics
  • long-format transformation
  • pheatmap clustering
  • scaled expression visualization
  • multi-panel figure construction
  • publication-style visualization
⬅️ Previous Exercise 🏠 Home ➡️ Next Exercise