Visualizing Gene Expression with Heatmaps in R

class: center, middle, inverse, title-slide

.title[
# Visualizing Gene Expression with Heatmaps in R
]
.author[
### Mikhail Dozmorov
]
.institute[
### Virginia Commonwealth University
]
.date[
### 2025-11-17
]

---

<style>
.large { font-size: 130%; }
.small { font-size: 70%; }
.tiny { font-size: 40%; }
</style>

## Introduction

* **Heatmaps** are graphical representations of data where individual values are represented as colors.
* Commonly used in genomics to visualize:

* Expression of genes across samples
  * Sample clustering patterns
  * Co-expression among genes

---

## The `airway` dataset

``` r
library(airway)
data(airway)
airway
```

* RNA-seq data of **human airway smooth muscle cells** treated with **dexamethasone** (a corticosteroid).

* 8 samples: 4 controls and 4 treated.

* Expression values are counts (from *summarizedExperiment*).

---

## DESeq2 analysis

We'll identify **differentially expressed genes** between treated and untreated samples.

``` r
dds <- DESeqDataSet(airway, design = ~ cell + dex)
dds <- DESeq(dds)
res <- results(dds)
res <- res[order(res$padj), ]
head(res)
```

---

## Select top 20 differential genes

``` r
top_genes <- rownames(res)[1:20]
mtx <- assay(vst(dds))[top_genes, ]  # variance-stabilized data
head(mtx)
```

---

## Row-centering

* Each gene may have different baseline expression.

* To focus on **relative expression changes**, we **center rows**: `$z_{ij} = x_{ij} - \bar{x}_{i}$`

``` r
mtx_centered <- t(scale(t(mtx), center = TRUE, scale = FALSE))
```

---

## Heatmap with `pheatmap`

``` r
annotation_col <- data.frame(
  Treatment = colData(dds)$dex,
  row.names = colnames(mtx_centered)
)

pheatmap(
  mtx,
  annotation_col = annotation_col,
  scale = "row",
  show_rownames = TRUE,
  show_colnames = TRUE,
  clustering_distance_rows = "euclidean",
  clustering_distance_cols = "euclidean",
  clustering_method = "complete",
  main = "Top 20 Differentially Expressed Genes (pheatmap)"
)
```

---

## Explanation

* **Row centering**: focuses on variation across samples rather than absolute expression.
* **Hierarchical clustering**:

* Groups **genes** with similar expression patterns.
  * Groups **samples** based on their expression profiles.
* **Annotations**: help visualize sample groups (e.g., treatment).

---

## Correlation-based Clustering

We can cluster genes using **(1 - Pearson correlation) / 2** as the distance measure
instead of the default Euclidean distance.

``` r
# Compute correlation-based distances
dist_rows <- as.dist((1 - cor(t(mtx), method = "pearson")) / 2)
dist_cols <- as.dist((1 - cor(mtx, method = "pearson")) / 2)

# Plot heatmap with correlation distances
pheatmap(
  mtx,
  annotation_col = annotation_col,
  scale = "row",
  show_rownames = TRUE,
  show_colnames = TRUE,
  clustering_distance_rows = dist_rows,
  clustering_distance_cols = dist_cols,
  clustering_method = "ward.D2",
  main = "Top 20 Differentially Expressed Genes (Correlation-based Clustering)"
)
```

---

## Interactive Heatmaps in R

- Interactive heatmaps allow **zooming**, **hover tooltips**, and **dynamic clustering**.
- Especially useful for **exploring large gene expression matrices**.

``` r
library(heatmaply)

heatmaply(
  mtx_centered,
  k_row = 3, k_col = 2,  # cluster highlights
  scale = "none",
  colors = colorRampPalette(c("blue", "white", "red"))(256),
  main = "Interactive Heatmap (heatmaply)",
  xlab = "Samples", ylab = "Genes",
  showticklabels = c(TRUE, TRUE)
)
```

---
## Heatmap with `ComplexHeatmap`

``` r
library(circlize)
col_fun <- colorRamp2(c(-2, 0, 2), c("blue", "white", "red"))

ComplexHeatmap::Heatmap(
  mtx_centered,
  name = "Expression",
  top_annotation = ComplexHeatmap::HeatmapAnnotation(Treatment = annotation_col$Treatment),
  col = col_fun,
  show_row_names = TRUE,
  show_column_names = TRUE,
  cluster_rows = TRUE,
  cluster_columns = TRUE,
  row_title = "Genes",
  column_title = "Samples",
  heatmap_legend_param = list(title = "Centered\nExpression")
)
```

---

## Comparison

| Feature       | `pheatmap`  | `ComplexHeatmap`                |
| ------------- | ----------- | ------------------------------- |
| Ease of use   | Very simple | More complex but powerful       |
| Customization | Moderate    | Extensive                       |
| Annotations   | Basic       | Rich (multi-layered)            |
| Integration   | Standalone  | Works well with tidyverse, grid |

---

## Summary

* **Heatmaps** reveal global expression patterns and clustering relationships.

* **Row centering** emphasizes relative variation.

* **Hierarchical clustering** groups similar genes and samples.

* Tools:

* `pheatmap`: simple and effective
  * `ComplexHeatmap`: advanced and customizable
  * `heatmaply`: interactive heatmaps