tools-only Heatmaps and Clustering - Usage Guide
Overview
Clustered heatmaps visualize matrix data (expression, methylation, etc.) with hierarchical clustering to reveal patterns across samples and features.
Prerequisites
# R
install.packages('pheatmap')
BiocManager::install('ComplexHeatmap')
# Python
pip install seaborn matplotlib scipy
Quick Start
Tell your AI agent what you want to do:
- "Create a heatmap of my top DE genes with sample annotations"
- "Cluster my expression matrix and extract cluster assignments"
- "Make a heatmap with row and column annotations"
Example Prompts
Basic Heatmaps
"Create a heatmap of my expression matrix with clustering"
"Make a quick heatmap of my top 50 variable genes"
Annotations
"Add sample annotations showing treatment and batch"
"Color row annotations by gene pathway"
Complex Layouts
"Split my heatmap by gene pathway"
"Create side-by-side heatmaps for two conditions"
What the Agent Will Do
- Prepare and scale the data matrix (typically z-score rows)
- Set up annotation data frames for rows/columns
- Choose appropriate distance metric and clustering method
- Configure color palette and legend
- Generate and export the heatmap
Tool Selection
| Tool | Language | Best For |
|---|---|---|
| pheatmap | R | Quick, publication-ready heatmaps |
| ComplexHeatmap | R | Complex annotations, multiple heatmaps |
| seaborn clustermap | Python | Python workflows, simple annotations |
Tips
- Scale data (z-score rows) for expression data
- Use correlation distance for expression, euclidean for abundance
- ward.D2 linkage often works well for biological data
- Use diverging palettes (RdBu) for centered data, sequential otherwise
- Limit to 50-100 genes for readability
Related Skills
- data-visualization/color-palettes - Color selection
- differential-expression/de-visualization - DE-specific plots