From MD Anderson Bioinformatics
| Next-Generation (Clustered) Heat Maps
|| Next-Generation (Clustered) Heat Maps are interactive heat maps that enable the user to zoom and pan across the heatmap, alter its color scheme, generate production quality PDFs, and link out from rows, columns, and individual heatmap entries to related statistics, databases and other information.
|| An NG-CHM server for creating your own test NG-CHMs is now available! Advanced users can now download a local copy of the NG-CHM system.
|| Discussion on Github
Examples: NG-CHM for The Cancer Genome Atlas (TCGA)
Our NG-CHM Pan-Cancer Compendium Portal contains numerous TCGA-related NG-CHM.
These NG-CHM are hosted on our
NG-CHM server, which provides the public with view-only access to many NG-CHMs.
Before using TCGA data, please read TCGA guidelines for publication and moratoriums.
We have a short Introductory Video describing the key features of the NG-CHM system. (Note: This video starts with 10 seconds of black and describes an earlier version of the system.) You can also watch the video on youtube.
The NG-CHM User Guide describes the NG-CHM user interface.
Next Generation Clustered Heat Maps
Heat maps are graphical representations that have been used for decades to present easily digested summaries of two-dimensional data matrices. Since 1993, they have been used for displaying biological data, and their use in displaying such data is now ubiquitous.
For high-throughput biological data, traditional static heat maps have several major drawbacks:
- The large size of high-throughput biological data (tens of thousands of variables per sample, or more, and hundreds to thousands of samples) limits static heat maps to showing either an overview of the overall pattern, or a detailed view of a tiny part of it.
- There is no mechanism for the user to interactively move from examining data elements to accessing related information. For small data sets with at most a few tens of variables or samples, this is a small inconvenience, but for large data sets it is a significant issue. For instance, identifying what characteristics a group of similar genes or samples have in common often requires the assistance of a specialized bioinformatician.
- The color scheme used in a static heat map is fixed when it's created.
- Comparing traditional heat maps of two independent variables is difficult.
Our conception of Next-Generation (Clustered) Heat Maps (NG-CHM) addresses these drawbacks:
- Interactive controls allow users to zoom and pan around a NG-CHM to see whatever parts of it they desire to, from its entirety through major regions of it, all the way down to a handful of adjacent data elements.
- Users can follow links from heat map elements (rows, columns, matrix elements, and/or groups of these) to related information or analyses. For instance, the user can select a cluster of co-regulated genes and follow a link to an analysis of the characteristics that these genes have in common.
- Users can interactively choose the color scheme that works best for them. This could be for aesthetic reasons (for instance, choosing a color scheme appropriate to ones color vision) or scientific ones, such as highlighting a specific range of values of interest.
- Users can juxtapose two or more independent variables, either temporally or by superposition, to identify potentially interesting relationships between them.
In addition, the NG-CHM system provides the other features you would expect in a full-featured clustered heat map solution, including the ability to generate publication-quality graphics.
Next-Generation (Clustered) Heat Map Components
Our implementation of the NG-CHM concept consists of multiple components:
- a NG-CHM Server, which serves the user interface elements to the user's browser on demand,
- a NG-CHM Java Builder, which creates a NG-CHM from a detailed specification,
- an R library, which provides an easy interface for a biostatistician or bioinformatician to use the NG-CHM Java builder in scripts,
- NG-CHM Comprehensive Web Builder, which provides an easy interface for an end-user to create a NG-CHM, and
- NG-CHM Quick Web Builder, which provides a very easy interface for an end-user (for instance, a biologist) to create a NG-CHM.
A compendium of read-only NG-CHMs showing The Cancer Genome Atlas (TCGA) data is available at http://bioinformatics.mdanderson.org/TCGA/NGCHMPortal/ .
A publicly accessible server for creating your own test NG-CHMs is available at http://bioinformatics.mdanderson.org/testchm .
Users with advanced system administrator skills can install a local NG-CHM system instance based on docker. Installation instructions are available at NG-CHM:Docker. Once installed, users can create NG-CHMs using the built-in Builder Web Application (similar to the one on our test server). Advanced bioinformatics analysts can also use
our high-level R library for building NG-CHMs.
In the near future, we are planning to release:
- a user-friendly system enabling ordinary users to install a local NG-CHM system.
- Bradley M. Broom
- Rehan Akbani
- John N. Weinstein
- In Silico
- Michael Ryan
- James Cleland
- SRA International
- David Kane
- Deepti Dodda
- Lam Nguyen
- Panna Shetty
- Sid Acharya
- MD Anderson
- Allen T. Chang
- Jun Zhang
- Sylvain Laroche
- TCGA: Grant number U24CA143883 from NCI/NIH
- The Michael & Susan Dell Foundation: The Lorraine Dell Program in Bioinformatics for Personalization of Cancer Medicine
- The H.A. Mary K. Chapman Foundation
- Anonymous donor for Computational Biology in Cancer Medicine