Gene expression heatmap across human tissues
Median TPM for 24 tissue marker genes across 12 human tissues, from Expression Atlas E-MTAB-5214 (CC BY 4.0), accessed 26 October 2026.
Heatmap GeneratorFigran
Paste a matrix, cluster rows and columns, and download the heatmap as SVG or PNG.
- CSV, TSV or Excel
- Clustering and z-scores
- Free SVG and PNG
Paste a numeric matrix. Your data stays in this browser.
One label and group per line, separated by a comma or Tab. Up to 8 groups per strip.
Grayscale cannot show sign for a diverging scale; show values or use a sequential scale.
Heatmap examples
Examples use the cited open datasets. Values marked illustrative are for demonstration only.
Correlation matrix heatmap of wine chemistry
A correlation heatmap of Pearson r for 13 chemical measurements across 178 samples, from UCI Wine dataset (CC BY 4.0), accessed 26 October 2026.
BLOSUM62 substitution matrix
A 20 × 20 amino acid substitution matrix heatmap, from NCBI BLOSUM62 matrix (public numerical data), accessed 26 October 2026.
Global temperature anomaly, year × month
A year × month table of temperature anomalies (°C) for 1980–2025, relative to 1951–1980, from NASA GISTEMP v4 (U.S. government data), accessed 26 October 2026.
Confusion matrix heatmap
Linear discriminant analysis on Fisher's 150 iris flowers, scored on the training data (resubstitution). Data: UCI Iris dataset (CC BY 4.0).
Illustrative values
Dose × time response matrix
Cell viability (%) across six doses and seven time points, with a custom color center of 100. Illustrative values.
How to make a heatmap
- Matrix · TSV
Gene Liver Pancreas Heart ALB 17866 316 0.5 APOA1 6111 0.8 11 TTR 1150 112 0.1
1 Prepare a matrix
Put row labels in the first column and column labels in the first row. Save as CSV or paste from Excel.
- Values + ClusteringLog2(x + 1)OnScaleRow z-scoreDistance · LinkageEuclidean · Complete
2 Transform and cluster
Use log2 and row z-scores for expression data, Pearson distance with average linkage, or the pheatmap defaults: Euclidean and complete.
- DownloadSVGheatmap.svg · vector text and cells
3 Choose a scale and download
Pick a colorblind-safe scale, add group strips, and download SVG or PNG. Report the distance, linkage and scaling in your figure legend.
Clustering, z-scores and color scales explained
Row z-score vs log scale
For expression data, log2(x + 1) compresses the dynamic range before z-score normalization. A row z-score compares each gene with its own mean and sample standard deviation, so it shows within-row patterns, not expression differences between genes.
Hierarchical clustering and the dendrogram
A heatmap with a dendrogram reorders rows and columns by similarity. Each leaf is one row or column; merge height shows the distance at which two clusters join. A cut separates the final groups.
| Distance | Euclidean: absolute differences. Pearson (1 − r): similarity of patterns. |
| Linkage | Complete favors compact clusters; average balances distances; Ward favors clusters with small within-group variance. |
| Report | Distance, linkage and transformation in the figure legend. |
Rotating a branch does not change the clustering; leaf order follows the R hclust convention so your figure matches pheatmap.
Leaf: each row or column · Cut: split the last merges into groups.
Choosing a heatmap color scale
Use a sequential scale such as Viridis for values moving in one direction. Use a diverging color scale such as RdBu when a center like zero has a meaning. The heatmap legend names the values and their range.
Avoid rainbow and red–green scales: perceived changes can distort the data or disappear for colorblind readers (Crameri et al., 2020). Viridis stays ordered in grayscale; RdBu needs values to distinguish positive and negative in black and white.
Heatmap FAQ
What is a heatmap in research?
A heatmap chart is a data visualization that maps each value in a matrix to a color. Rows and columns can be reordered by hierarchical clustering, with dendrograms showing similarities. Gene expression, correlation matrices, confusion matrices and time × condition tables are common uses.
What format does my data need?
Paste CSV or TSV, or upload an XLSX workbook. Use the first row and column for labels and a dot for decimals. Blank cells and NA are missing values. You can copy directly from Excel. The limit is 40,000 cells, at most 2,000 rows or 200 columns. SVG keeps cells and text editable in Illustrator, Inkscape or PowerPoint.
Should I use row z-scores or log values?
For counts or TPM, log2(x + 1) compresses the range before row z-scores. Each row is centered and divided by its sample standard deviation, which compares patterns within a gene. It does not compare absolute expression across genes. Keep correlation and confusion matrices unscaled.
Which distance and linkage should I choose?
The clustering is agglomerative: it starts from single rows or columns and repeatedly joins the two closest clusters. Euclidean measures absolute differences; Pearson distance compares patterns. Complete favors compact clusters, average (UPGMA) balances distances, and Ward (ward.D2) minimizes the increase in within-cluster variance and is conventionally paired with Euclidean distance. There is no universally correct choice: report it in the caption. Single linkage is not offered because it can chain clusters together.
Will the clustering match pheatmap or ComplexHeatmap?
The defaults are Euclidean distance and complete linkage, matching pheatmap and ComplexHeatmap. Ward uses ward.D2. Clustering follows transformation, with R hclust merge, tie and leaf-order conventions tested against R 4.6.0 and pheatmap 1.0.13, including tied data. Missing values differ: correlation uses pairwise-complete observations and constant rows get z-scores of zero with a warning, where R can produce NaN or an error.
Which color scale is colorblind-safe and prints in grayscale?
Viridis, Magma and Blues are colorblind-safe with ordered luminance, so grayscale preserves their order. RdBu and PuOr are also colorblind-safe, but grayscale cannot show the sign on either side of their center. Show values or choose a sequential scale for black-and-white printing. Rainbow and red–green scales are not offered.
Can I make a heatmap from Excel data?
Upload .xlsx to read the first worksheet or paste cells from Excel. An Excel heat map made with conditional formatting has no hierarchical clustering, dendrograms or heatmap legend. Download SVG to retain editable text for your paper.
How is this different from Heatmapper, Morpheus or seaborn clustermap?
Heatmapper is a free server-based web tool for several heatmap types (Babicki et al., 2016). Morpheus is Broad’s web tool for large matrices and interactive exploration. seaborn clustermap (Python, built on matplotlib) and pheatmap (R) require code and offer extensive customization. This tool renders pasted data in your browser and downloads SVG or PNG, without uploading your matrix. It does not perform statistical analysis.
Is this heatmap maker free, and is my data private?
Yes. The online heatmap tool is free and needs no login. Parsing, clustering, rendering and downloads run in your browser. Matrix data is not uploaded or included in product events. Drafts stay in this tab’s sessionStorage.
Ready to draw your heatmap?
Paste your matrix or load an example, then download the figure.