version license

Clustergrammer-JS is the front end JavaScript library that builds the interactive clustergram visualization in SVG using the visualization library D3.js. The library is free and open-source and can be found on GitHub.

Clustergrammer-JS Dependencies


Clustergrammer.js can be installed using node package manager (npm package) with the following:

npm install clustergrammer

or the source code and library, clustergrammer.js, can be obtained from the Clustergrammer GitHub repo.

JavaScript Workflow Example

This workflow shows how to generate a visualization using the JSON produced by

// the visualization JSON (produced by Clustergrammer-PY)
var network_data = {
                      "mat": [...]

// args must contain root of container and the visualization JSON
var args = {
  'root': '#id_of_container',
  'network_data': 'network_data'

// Clustergrammer returns a Clustergrammer object in addition to making
// the visualization
var cgm = Clustergrammer(args);

The id of the container where the visualization will be made is passed as root (this root container must be made by the user). The Visualization-JSON (produced by Clustergrammer-PY) contains all the information necessary to generate the visualization and is passed in this example as network_data. See the Clustergrammer-JS API for additional arguments that can be passed to Clustergrammer.js.

Example Pages

The Clustergrammer GitHub repo contains several example pages demonstrating how to build a web page with a Clustergrammer visualization. The page index.html and corresponding script load_clustergram.js show how to make a full-screen resizable visualization. The page multiple_clust.html and corresponding script load_multiple_clustergrams.js show how to visualize multiple clustergrams on one page. Note that each visualization requires its own container.

Clustergrammer-JS API

class Clustergrammer(args)

The Clustergrammer JavaScript object takes the args object and produces a visualization on the page.

This args object has two required arguments, network_data and root:


This required attribute is where the visualization JSON should be passed as a JavaScript object.


This required attribute is the id (passed as a string) of the container where Clustergrammer will be built. Each Clustergrammer visualization in a page should be passed a unique id.

Clustergrammer.args.row_label, args.col_label

Pass strings that will be used as ‘super-labels’ for rows and columns.

Clustergrammer.args.row_label_scale, args.col_label_scale.

Scaling factor to increase/decrease the size of the rows and column labels.


Scaling factor to increase/decrease the size of the row/column ‘super-labels’.


Name of the scaling function, e.g. linear, log, used to map matrix cell values to cell opacity. The default is linear.


The input_domain defines the maximum absolute value of matrix cells that are mapped to the maximum opacity of 1. The default input_domain is defined using the maximum absolute value of the matrix. Lowering the input_domain value increases the opacity of the overall visualization by setting a cutoff.


This boolean value turns on or off zooming. The default is true.


Set the positive and negative colors in the heatmap using an array with color names or hexcode, e.g. ['#ED9124','#1C86EE']. The default is red and blue for positive and negative, respectively.

Clustergrammer.args.row_order, args.col_order

Set the initial ordering for rows and columns. The default is clust and the options are:

  • alpha: order based on names
  • clust: order based on clustering
  • rank: order based on sum
  • rank_var: order based on variance

Load clustergram using an initial filtered view.


This attribute is a string (which can include HTML) that will produce a small About section at the top of the sidebar. This can be used to provide a quick description about the data or visualization.


Users can pass a callback function that will run when mousing over row labels.


Users can pass a callback function that will run when mousing over column labels.


Users can pass a callback function that will run when mousing over a matrix-cell (e.g. matrix tile).


Users can pass a callback function that will run when mousing over a dendrogram cluster (e.g. gray trapezoid)


Users can pass a callback function that will run when clicking a dendrogram cluster (e.g. gray trapezoid)


Users can pass a callback function that will run anytime the matrix has been updated, for instance when filtering/un-filtering, cropping, etc.


Initialize the visualization in ‘expanded’ mode where the sidebar is not visible. The sidebar can be shown by clicking the menu button on the top left of the visualization.


Users can modify the width of the sidebar by specifying the width of the sidebar in pixels as a number.


Users can initialize the ‘view’ of their matrix, e.g. initialize the matrix at a particular row filtering level.


This boolean option gives users have the option to not make any Bootstrap modals (e.g. dendrogram group modals) and the default is true.

Clustergrammer’s attributes and functions are listed below:


The Clustergrammer parameters object, which contains all the parameters necessary to generate the visualization.


Update the visualization row categories.

  • row_data – Row category data.

Reset the row categories to their original state.


Call this function to resize the visualization to fit in its resized container (if the user has resized the container).


Generate a D3 tooltip for SVG elements.

Clustergrammer.update_view(filter_type, inst_state)

Update the heatmap with a specified row filter ‘view’.

  • filter_type – The available filter types sum or variance: e.g. N_row_sum, N_row_var
  • inst_state – The value of the row filter, e.g. 500

Update the visualization to show the row and column names specified in the names object.

  • names – Object with row and/or col attributes that specify the row and column names that will be visible after updating. Row and column names should be given as a list. Users can include only one attribute, e.g. filter rows only by including no col attribute, to only filter rows or columns (or users can specify an empty list to not filter).

Update the visualization to show the row and column names specified in the nodes object.

  • names – Object with row and col attributes that specify the row and column nodes that will be visible after updating.
Clustergrammer.zoom(pan_x, pan_y, zoom)

Zoom and pan into the visualization.

  • pan_x – Panning in the x direction
  • pan_y – Panning in the y direction
  • zoom – The zoom level applied to the visualization.

Save the current matrix (e.g. after cropping) as a tab-separated file.


The visualization-JSON is calculated by Clustergrammer-PY and encodes everything needed for the front end Clustergrammer-JS to produce the visualization. The visualization-JSON format is described here (see clustergrammer_example.json for an example file). An overview of the format is shown below (note that the group arrays are not shown):

      "name": "ATF7",
      "clust": 67,
      "rank": 66,
      "rankvar": 10,
      "group": []
      "name": "Col-0",
      "clust": 4,
      "rank": 10,
      "rankvar": 120,
      "group": []
    [1, 2],
    [3, 4],
    [5, 6]
      "source": 0,
      "target": 0,
      "value": 0.023

Optional ‘views’ of the matrix (e.g. row-filtered views) are encoded into the views attribute at the base level of the object. These views are used to store a filtered version of the matrix. Only the row and column names are stored in these views since all views share the same matrix cells. The view attributes are stored in the view object (e.g. N_row_sum):

    "N_row_sum": "all",
    "dist": "cos",
      "row_nodes": [],
      "col_nodes": []

There are three required properties for the Visualization-JSON: row_nodes, col_nodes, and mat (links can be used in place of mat and will continue to be supported, but the default format will use mat). Each of these properties is an array of objects and these objects are discussed below.


row_nodes and col_nodes objects are required to have three properties: name, clust, rank. name specifies the name given to the row or column. clust and rank give the ordering of the row or column in the clustergram. Two optional properties are group and value. group is an array that contains group-membership of the row/column at different dendrogram distance cutoffs and is necessary for displaying a dendrogram. If nodes have the value property, then semi-transparent bars will be displayed behind the labels to represent this value.


mat is an JavaScript array that stores the matrix data. The source and target of each value (row and column) are inferred from the position of the data in the two-dimensional array.


Note: mat will be used by default instead of links, but both formats will be supported (mat is usually a more compact format). links have three properties: source, target, and value. source and target give the integer value of the row and column of the matrix-cell in the visualization. value specifies the opacity and color of the matrix-cell, where positive/negative values results in red/blue matrix-cells in the visualization. The optional properties value_up and value_dn allow the user to have a split matrix-cell that has an up-triangle and a down-triangle.

Users can also generate the visualization-JSON using their own scripts provided that they adhere to the above format.

Clustergrammer-JS Development

The Clustergrammer-JS source code can be found in the Clustergrammer GitHub repo. The Clustergrammer-JS library is utilized by the Clustergrammer-Web and the Clustergrammer-Widget. Clustergrammer-JS is built with Webpack Module Bundler from the source files in the src directory.

Please Contact Nicolas Fernandez and Avi Ma’ayan with questions or use the GitHub issues feature to report an issue.