Exercise 9 — Interpreting DESeq2 Results and Prioritizing Candidate Genes

In Exercise 8, we performed a full differential expression analysis using DESeq2.

We now have a table containing:

  • log2FoldChange
  • p-value
  • adjusted p-value
  • mean expression

But a common challenge in RNA-seq analysis is:

We have thousands of significant genes. Which ones matter?

In this exercise, you will learn how to:

  • interpret DESeq2 results
  • prioritize genes
  • create publication-style volcano plots
  • identify candidate biomarkers
  • think critically about differential expression results

Part 1 — Load Previous Results

Use your DESeq2 object from Exercise 8.

res <- results(dds)
res_df <- as.data.frame(res)

Part 2 — Understanding DESeq2 Output

Task 2.1

Inspect:

head(res_df)

Answer:

  1. What does log2FoldChange represent?
  2. What does pvalue represent?
  3. What does padj represent?
  4. Why is padj preferred over pvalue?

Task 2.2

Interpret the following hypothetical genes:

Gene A

log2FC = 3
padj = 0.001

Questions:

  • Upregulated or downregulated?
  • Strong or weak effect?
  • Significant or not?

Gene B

log2FC = 0.2
padj = 1e-12

Questions:

  • Significant?
  • Biologically interesting?
  • Why or why not?

Gene C

log2FC = 2.5
padj = 0.4

Questions:

  • Large effect?
  • Reliable finding?
  • Would you prioritize it?

Part 3 — Significant Genes

Task 3.1

Identify:

padj < 0.05

How many genes are significant?

Task 3.2

Identify:

padj < 0.05 &
abs(log2FoldChange) > 1

How many genes pass both criteria?

Task 3.3

Why might researchers use both thresholds?

Part 4 — Ranking Candidate Genes

Task 4.1

Sort by:

padj

Display:

Top 20 genes

Task 4.2

Sort by:

abs(log2FoldChange)

Display:

Top 20 genes

Task 4.3

Compare both lists.

Questions:

  1. Are they the same?
  2. Why not?
  3. Which ranking is more useful?

Part 5 — Volcano Plot

Task 5.1

Create:

volcano_df <- res_df

Task 5.2

Add significance label:

volcano_df$Significant <- ifelse(
    volcano_df$padj < 0.05,
    "Yes",
    "No"
)

Task 5.3

Create volcano plot

library(ggplot2)

ggplot(
    volcano_df,
    aes(
        x = log2FoldChange,
        y = -log10(padj),
        color = Significant
    )
) +
    geom_point()

Task 5.4

Interpret:

  1. What is the x-axis?
  2. What is the y-axis?
  3. What does the upper-right corner represent?
  4. What does the upper-left corner represent?

Part 6 — Improved Volcano Plot

Task 6.1

Add threshold lines:

padj = 0.05

and

log2FC = ±1

using:

geom_hline()
geom_vline()

Task 6.2

Questions:

  1. Which genes would you prioritize?
  2. Which region contains strongest candidates?

Part 7 — Candidate Biomarker Selection

Suppose your PI asks:

Give me 5 genes to validate experimentally.

Task 7.1

Create:

candidate_genes

using:

padj < 0.05

and

abs(log2FoldChange) > 2

Task 7.2

Select your top 5 candidates.

Task 7.3

Justify each choice.

Consider:

  • significance
  • fold change
  • expression level

Part 8 — Mean Expression Matters

A common mistake is selecting genes based only on fold change.

Task 8.1

Add:

baseMean

to your ranking strategy.

Task 8.2

Questions:

Why might a gene with:

log2FC = 4

be difficult to validate if:

baseMean = 2

?

Part 9 — Gene Prioritization Table

Create a summary table containing:

  • Gene
  • baseMean
  • log2FoldChange
  • padj

Sort by:

padj

Display top 20 genes.

Part 10 — Analyst Interpretation

Answer:

# What is the difference between statistical significance and biological significance?

# Why is fold change important?

# Why is adjusted p-value important?

# What makes a strong biomarker candidate?

# Which genes would you prioritize for validation?

Part 11 — Mini Report

Write 5–10 sentences summarizing:

  • Number of significant genes
  • Major findings
  • Volcano plot observations
  • Candidate genes
  • Recommended validation targets

Imagine presenting these results at a lab meeting.

Bonus Challenge

B1

Label the top 10 genes on the volcano plot.

Hint:

library(ggrepel)

B2

Colour genes using:

  • Upregulated
  • Downregulated
  • Not significant

instead of simple Yes/No.

B3

Create a publication-style volcano plot.

Add:

  • clean theme
  • labels
  • thresholds
  • title

⬅️ Previous Exercise | 🏠 Home | ➡️ Next Exercise