Email updates

Keep up to date with the latest news and content from BMC Bioinformatics and BioMed Central.

Open Access Highly Accessed Software

NeatMap - non-clustering heat map alternatives in R

Satwik Rajaram123* and Yoshi Oono12

Author Affiliations

1 Department of Physics, 1110 W. Green Street, University of Illinois at Urbana-Champaign, Urbana, IL 61801-3080, USA

2 Institute for Genomic Biology, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA

3 Current address: Department of Pharmacology, University of Texas Southwestern Medical Center, Dallas, TX 75390, USA

For all author emails, please log on.

BMC Bioinformatics 2010, 11:45  doi:10.1186/1471-2105-11-45

Published: 22 January 2010

Abstract

Background

The clustered heat map is the most popular means of visualizing genomic data. It compactly displays a large amount of data in an intuitive format that facilitates the detection of hidden structures and relations in the data. However, it is hampered by its use of cluster analysis which does not always respect the intrinsic relations in the data, often requiring non-standardized reordering of rows/columns to be performed post-clustering. This sometimes leads to uninformative and/or misleading conclusions. Often it is more informative to use dimension-reduction algorithms (such as Principal Component Analysis and Multi-Dimensional Scaling) which respect the topology inherent in the data. Yet, despite their proven utility in the analysis of biological data, they are not as widely used. This is at least partially due to the lack of user-friendly visualization methods with the visceral impact of the heat map.

Results

NeatMap is an R package designed to meet this need. NeatMap offers a variety of novel plots (in 2 and 3 dimensions) to be used in conjunction with these dimension-reduction techniques. Like the heat map, but unlike traditional displays of such results, it allows the entire dataset to be displayed while visualizing relations between elements. It also allows superimposition of cluster analysis results for mutual validation. NeatMap is shown to be more informative than the traditional heat map with the help of two well-known microarray datasets.

Conclusions

NeatMap thus preserves many of the strengths of the clustered heat map while addressing some of its deficiencies. It is hoped that NeatMap will spur the adoption of non-clustering dimension-reduction algorithms.