How to Make a Heatmap for Scientific Papers
A complete guide to creating publication-quality heatmaps. From data normalization to journal-ready export.
What is a scientific heatmap?
A heatmap is a data visualization that uses color intensity to represent values in a matrix. In scientific papers, heatmaps are used for gene expression data, correlation matrices, protein interaction networks, and any large dataset where patterns need to be revealed visually.
Key requirements:
- • Normalized data (z-score, log2, or min-max)
- • Appropriate color scale (diverging or sequential)
- • Row and column labels (concise for large matrices)
- • Color scale bar with units
- • Clustering method stated in legend (if applicable)
- • Journal width (single: 84–90 mm, double: 170–183 mm)
- • 300 DPI minimum for raster export
Step-by-Step Guide
- Prepare your data matrix
Organize data with rows (genes, proteins, samples) and columns (conditions, time points, treatments). Ensure values are numeric and consistent.
- Normalize values
Apply normalization: z-score for centered data, log2 for fold changes, min-max for bounded data. Normalization ensures fair comparison across rows and columns.
- Choose color scale
Diverging (blue-white-red) for centered data. Sequential (white-to-red) for magnitude. Ensure colorblind-friendly and print-compatible.
- Add labels
Label rows and columns concisely. For large matrices, show only key rows or use clustering. Add a color scale bar with units.
- Cluster if needed
Apply hierarchical clustering to rows and/or columns to reveal patterns. State the method and distance metric in the legend.
- Export for publication
Set width to journal column or double column. Export at 300 DPI. Ensure text is readable at journal size. Use vector format if possible.
Color Scale Selection
| Data Type | Color Scale | Example |
|---|---|---|
| Z-scores, log2 ratios | Diverging (blue-white-red) | Gene expression |
| Counts, percentages | Sequential (white-to-red) | Protein abundance |
| Correlation (-1 to +1) | Diverging (blue-white-red) | Correlation matrix |
| P-values | Sequential (white-to-red) | Significance matrix |
Frequently Asked Questions
How do you make a heatmap for a scientific paper?
To make a heatmap: (1) organize your data matrix (rows = genes/samples, columns = conditions/time points), (2) normalize values (z-score, log2, or min-max), (3) choose a color scale (diverging for centered data, sequential for magnitude), (4) add row and column labels, (5) include a color scale bar, (6) cluster rows/columns if showing relationships, (7) export at 300 DPI at journal width.
What is a heatmap used for in science?
Heatmaps are used to visualize large matrices of data where color intensity represents value magnitude. Common uses include gene expression data, correlation matrices, protein interaction networks, and time-course data. Heatmaps reveal patterns, clusters, and outliers that are hard to see in tables.
What color scale should I use for heatmaps?
Use a diverging color scale (e.g., blue-white-red) for data centered around zero (z-scores, log ratios). Use a sequential scale (e.g., white-to-red) for magnitude-only data (counts, percentages). Ensure the scale is colorblind-friendly and print-compatible.
What is the best tool for making heatmaps?
FigureGuild is ideal for publication-ready heatmaps. It auto-normalizes data, applies color scales, and exports at journal dimensions. R (pheatmap, ComplexHeatmap) and Python (seaborn, matplotlib) are also used but require coding. GraphPad Prism does not support heatmaps natively.
Should heatmaps show row and column labels?
Yes, but only essential labels. For large matrices (100+ rows), show only key rows or use row clustering with dendrograms. Column labels should be concise. Consider providing a separate table with full labels as supplementary data.
How do you cluster rows and columns in a heatmap?
Clustering organizes rows and columns by similarity. Hierarchical clustering (average linkage, Euclidean distance) is most common. Clustering reveals groups of genes with similar expression patterns or samples with similar profiles. Always state the clustering method and distance metric in the figure legend.
Related Pages
Create Heatmaps Automatically
FigureGuild auto-normalizes data, applies color scales, and exports at journal dimensions — all from your raw matrix.
Try FigureGuild Free