“This paper surveys biclustering algorithms, a technology used in the analysis of gene expression data.”

This paper surveys biclustering algorithms, a technology used in the analysis of gene expression data. This technology improves the quality and detail of the knowledge that can be gathered from microarray experiments. Since this high throughput technique has become an essential tool for all molecular biologists in their quest to understand the complex mechanisms that control all biological systems, scientists from a number of disciplines have shown a strong interest in this area. We consider that this particular survey has been highly cited because it presents the state of the art in a systematic, organized way, which makes it easy to understand and classify existing approaches.

This paper represents essentially a synthesis of existing knowledge, in the area of biclustering algorithms. This synthesis is presented in a systematic way, and uses a number of dimensions of analysis to classify the methods and algorithms that have been proposed to date.

A microarray experiment obtains, in a first approximation, the levels of expression of a large set of genes in an organism, in a specific experimental condition. Analyzing data from multiple microarray experiments is a daunting task that requires the use of computer algorithms to process and visualize the data.

One of the methods most commonly used groups the genes by their global patterns of expression. This enables scientists to study large groups of genes that have a similar behavior. However, grouping genes by their global expression levels ignores the important fact that regulation mechanisms lead genes to behave in different ways, depending on the activation of the different processes where they are involved.

Biclustering provides a mechanism to analyze the behavior of groups of genes in a more detailed way, by finding groups of genes that have highly correlated behaviors in a particular subset of experimental conditions. Such a group of genes is likely to be involved in a specific biological process in those conditions. This information can be used to better understand biological processes, as well as the relationships between genes.

This line of research is included in the general strategy of developing tools and methods that will advance the discipline of systems biology. Such tools are being developed in a joint effort of a number of research groups that work in the areas of computer science, data analysis, and biological sciences. The major challenges are related with the complexity of the problem being addressed, and the need for close cooperation between interdisciplinary groups. These factors, together with the importance of the problem, also represent major opportunities.

The specific techniques addressed in this article will prove useful in many areas, including high-profile fields like cancer research, biotechnology, and systems biology. Therefore, improving the methods used for the analysis of gene expression data can have a high social impact, by improving our living conditions. This impact, however, will be mostly indirect, since these techniques are instrumental in creating the knowledge that is required, but do not create that knowledge by themselves.