Clustering in life sciences.

Clustering is the task of organizing a set of objects into meaningful groups. These groups can be disjoint, overlapping, or organized in some hierarchical fashion. The key element of clustering is the notion that the discovered groups are meaningful. This definition is intentionally vague, as what constitutes meaningful is to a large extent, application dependent. In some applications this may translate to groups in which the pairwise similarity between their objects is maximized, and the pairwise similarity between objects of different groups is minimized. In some other applications this may translate to groups that contain objects that share some key characteristics, even though their overall similarity is not the highest. Clustering is an exploratory tool for analyzing large datasets, and has been used extensively in numerous application areas. Clustering has a wide range of applications in life sciences and over the years has been used in many areas ranging from the analysis of clinical information, phylogeny, genomics, and proteomics. For example, clustering algorithms applied to gene expression data can be used to identify co-regulated genes and provide a genetic fingerprint for various diseases. Clustering algorithms applied on the entire database of known proteins can be used to automatically organize the different proteins into closeand distant-related families, and identify subsequences that are mostly preserved across proteins [52, 22, 55, 68, 49]. Similarly, clustering algorithms applied to the tertiary structural datasets can be used to perform a similar organization and provide insights in the rate of change between sequence and structure [20, 65]. The primary goal of this chapter is to provide an overview of the various issues involved in clustering large datasets, describe the merits and underlying assumptions of some of the commonly used clustering approaches, and provide insights on how to cluster datasets arising in various areas within life-sciences. Toward this end, the chapter is organized in broadly three parts. The first part (Sections 2– 4) describes the various types of clustering algorithms developed over the years, the various methods for computing the similarity between objects arising in life sciences, and methods for assessing the quality of the clusters. The second part (Section 5) focuses on the problem of clustering data arising from microarray experiments and describes some of the commonly used approaches. Finally, the third part (Section 6) provides a brief introduction to CLUTO, a general purpose toolkit for clustering various datasets, with an emphasis on its applications to problems and analysis requirements within life sciences.