Exercise 5 — Differential Expression Analysis Basics

In RNA-seq analysis, one of the most common goals is to identify genes that behave differently between conditions.

Examples:

  • healthy vs disease
  • treated vs untreated
  • responder vs non-responder

This process is called:

Differential Expression (DE) Analysis

In this exercise, we will:

  • combine expression matrices with metadata
  • compare Control vs Treatment samples
  • calculate fold changes
  • identify candidate DE genes

This exercise introduces the core logic behind tools such as:

  • DESeq2
  • edgeR
  • limma

🧬 Biological Context

Suppose we performed RNA-seq on:

  • 4 Control samples
  • 4 Treatment samples

We want to identify genes that increase or decrease after treatment.

Part 1 — Create Expression Matrix

Task 1.1

Create the following gene vector:

genes <- c(
  "TP53","EGFR","MYC","BRCA1",
  "ACTB","GAPDH","VEGFA","CDKN1A",
  "STAT3","BCL2"
)

Task 1.2

Create the following count matrix:

counts <- matrix(c(

  120,130,125,128,220,230,240,235,
  300,310,295,305,500,520,510,530,
  90,85,88,92,70,72,68,65,
  450,470,460,455,700,720,710,730,
  1000,980,1020,1010,1050,1070,1080,1060,
  900,920,910,905,930,950,940,945,
  140,150,145,148,320,340,330,335,
  200,210,205,198,410,430,420,425,
  180,190,185,188,350,360,355,370,
  500,490,510,505,300,290,310,305

), nrow = 10, byrow = TRUE)

Task 1.3

Assign:

  • row names = genes
  • column names:
Control_1
Control_2
Control_3
Control_4
Treatment_1
Treatment_2
Treatment_3
Treatment_4

Part 2 — Create Metadata

Task 2.1

Create a metadata data frame containing:

Sample Condition
Control_1 Control
Control_2 Control
Control_3 Control
Control_4 Control
Treatment_1 Treatment
Treatment_2 Treatment
Treatment_3 Treatment
Treatment_4 Treatment

Store as:

metadata

Task 2.2

Inspect:

  • dimensions
  • row names
  • column names

Part 3 — Explore Expression Matrix

Task 3.1

Check:

  • dimensions of counts matrix
  • number of genes
  • number of samples

Task 3.2

Retrieve:

  • TP53 expression values
  • VEGFA expression values
  • STAT3 expression in Treatment_4

Part 4 — Calculate Mean Expression

Task 4.1

Calculate:

  • mean Control expression per gene
  • mean Treatment expression per gene

Store as:

control_mean
treatment_mean

Task 4.2

Combine results into a data frame:

de_results

containing:

  • Gene
  • ControlMean
  • TreatmentMean

Part 5 — Fold Change

Task 5.1

Calculate fold change:

TreatmentMean / ControlMean

Add as a new column:

FoldChange

Task 5.2

Interpret:

  • FoldChange > 1
  • FoldChange < 1

Task 5.3

Find genes with:

  • FoldChange > 1.5

These may represent:

  • upregulated genes

Task 5.4

Find genes with:

  • FoldChange < 0.8

These may represent:

  • downregulated genes

Part 6 — Ranking Genes

Task 6.1

Sort genes:

  • highest FoldChange → lowest

Task 6.2

Identify:

  • top 3 upregulated genes
  • top 3 downregulated genes

Part 7 — Expression Filtering

Genes with very low expression are often removed before DE analysis.

Task 7.1

Calculate:

  • mean expression across ALL samples

Task 7.2

Filter out genes where:

  • mean expression < 100

Store as:

filtered_de_results

Task 7.3

How many genes were removed?

Part 8 — Log2 Fold Change

RNA-seq analysis commonly uses:

log2 Fold Change

instead of regular fold change.

Task 8.1

Calculate:

log2FC = log2(FoldChange)

Add as a new column.

Task 8.2

Interpret:

  • positive log2FC
  • negative log2FC
  • log2FC = 0

Task 8.3

Find genes with:

  • log2FC > 0.5

Part 9 — Biological Interpretation

Answer in comments:

# 1. Which genes appear strongly upregulated?
# 2. Which genes appear downregulated?
# 3. Why is fold change useful?
# 4. Why do we use log2 fold change?
# 5. Why is metadata important in RNA-seq analysis?

Bonus Challenge

B1

Create a column:

Regulation

Rules:

  • “Upregulated” if log2FC > 0.5
  • “Downregulated” if log2FC < -0.5
  • “Stable” otherwise

B2

Count:

  • number of upregulated genes
  • number of downregulated genes
  • number of stable genes

B3

Find genes that are:

  • Upregulated AND
  • mean expression > 300

🧠 Key Concepts Learned

  • metadata-driven analysis
  • condition comparison
  • fold change
  • log2 fold change
  • gene ranking
  • candidate differential expression

These are the conceptual foundations behind:

  • DESeq2
  • edgeR
  • limma differential expression workflows
⬅️ Previous Exercise 🏠 Home ➡️ Next Exercise