node package manager


This is a clustergram implemented in D3.js. I started from the example and added the following features



Clustergrammer is an interactive web-based tool for visualizing high-dimensional data as heatmaps. The front-end JavaScript library, clustergrammer.js is built using D3.js and 'back-end' calculations are done using the Python library Click the screenshot below to view an interaceive tutorial:


The project began as an extension of this example and some of clustergrammer's interacive features include:

  • zooming/panning
  • multiple reordering options (e.g. variance)
  • interactive dendrogram
  • row filtering and searching
  • row/column categories

Clustergrammer can be used as:

Clustergrammer is designed to be a reusable chart and has been integrated into several Ma'ayan lab web tools including:

Using Clustergrammer

Clustergrammer consists of two parts: 1) the front-end JavaScript library clustergrammer.js used the make the interactive visualization, and 2) the 'back-end' used to cluster your data and make the JSON for the front-end visualization.

You can install clustergrammer.js by downloading the latest release or with npm npm install clustergrammer. You also install by downloading the latest release or with pip pip install clustergrammer.

The easiest way to visualize a matrix of your own data is to follow the example Python workflow that demonstrates how to use the Python library Users can also generate the JSON for clustergrammer.js (see the example mult_view.json) using their own scripts as long as they adhere to the format.

Clustergrammer JavaScript Library

Clustergrammer.js uses the visualization library D3.js to build an interactive heatmap visualization made using SVG. The Clustergrammer.js source code is under the src directory and Webpack Module Developer is being used to make clustergrammer.js.

To make a visualization pass an arguments object with the following required values to Clustergrammer:

// load visualization JSON to network_data
var args = {
  'network_data': network_data
var cgm = Clustergrammer(args);

The id of the container where the visualization SVG will be placed is passed as root (this root container must be made by the user). The visualization JSON (example here mult_view.json and format discussed here) contains the information necessary to make your visualization and is passed as network_data. The visualization JSON is produced by See additional optional clustergrammer.js arguments for additional options that can be passed to clustergrammer.js.

Clustergrammer.js Dependencies

  • D3.js
  • jQuery
  • Underscore.js

Visualization Resizing

The visualization can be resized by: first resizing the container and then resizing the visualization using cgm.resize_viz(). An example of resizing when the window change size is shown below.'resize', function(){
  // first, resize the container when the screen size changes
  // then, resize the visualization to match the container's new size

Example Webpages

The page index.html (and the corresponding script load_clustergram.js) demonstrates how to make a full-screen resizable clustergrammer visualization.

The page multiple_clust.html (and corresponding script load_multiple_clustergrams.js) demonstrates how to visualize multiple clustergrams on one page. Note that each visualization requires its own container.

Clustergrammer Python Library

The Clustergrammer python library, takes a tab-separated matrix file as input (see format here), calculates clustering, and generates the visualization json (see format here) for clustergrammer.js. The library can be installed using pip and is compatable with Python 2.7 and 3.5:

# Python 2
$ pip install clustergrammer
# Python 3
$ pip3 install clustergrammer

or the source code can be obtained from repo: simply copy the clustergrammer directory with the source code to the main directory to use the library in this repo. Dependencies

  • Numpy
  • Scipy
  • Pandas

Example Python Workflow

from clustergrammer import Network
net = Network()
# load matrix file
# calculate clustering
net.make_clust(dist_type='cos',views=['N_row_sum', 'N_row_var'])
# write visualization json to file
net.write_json_to_file('viz', 'json/mult_view.json')

The script is used to generate the visualization jsons for the examples pages in this repo. To visualize your own data modify the script.

Input Matrix Format discussed here takes a tab separated matrix with unique row and column names as input. The simplest format is shown here (note: that tabs are required, but spaces are used in the example below to increase readability):

       Col-A   Col-B   Col-C
Row-A   0.0    -0.1     1.0
Row-B   3.0     0.0     8.0
Row-C   0.2     0.1     2.5

Row and column categories can also be included in the matrix in the following way:


This screenshot of an Excel spreadsheet shows a single row category being added as an additional column of strings (e.g. 'Type: Interesting') and a single column category being added as an additional row of strings (e.g. 'Gender: Male'). Up to 15 categories can be added in a similar manner. Titles for row/column names or categories can be added by prefixing each string with Title:(note a space is required after the colon). For example the title of the column names is Cell Line and the title of the row categories is Gender .

Alternatively, row/column names and categories can be stored as Python tuples as shown below (or see tuple_cats.txt).

  ('Cell Line: A549', 'Gender: Male') ('Cell Line: H1299', 'Gender: Female')  ('Cell Line: H661', 'Gender: Female')
('Gene: EGFR','Type: Interesting')  -3.234  5.03  0.001
('Gene: TP53','Type: Not Interesting')  8.3 4.098 -12.2
('Gene: IRAK','Type: Not Interesting')  7.23  3.01  0.88

This format is easier to export from a Python Pandas DataFrame (see net.write_matrix_to_tsv). Note that 'titles' have been added to row and column names as well as row and column categories.

Several example tab-separated matrix files can be found in the txt directory. See example Python workflow or for examples of how to use to generate a visualization json from these matrix files.


Clustergrammer was developed by Nicolas Fernandez in the Ma'ayan lab at the Icahn School of Medicine at Mount Sinai. Clustergrammer's license and third-party licenses are in the LICENSES directory.