Clusteringanddisplaying.PDFVIP

Clustering and displaying microarray data George Bell, Ph.D. Bioinformatics and Research Computing Hot Topics – March 2008 Why? • Explore a large amount of expression or other data • Get experiment-wide look at interesting subset of data • Visually identify patterns for further analysis • Order genes and/or experiments in a sensible way • Split genes and/or experiments into a predefined number of groups Why not? • Clustering is not a substitute for rigorous statistics • Clustering cannot identify – differentially expressed genes – profiles that are correlated with a reference profile • Any data – even noise – can be clustered • Clustering is not an essential step for most analyses Where to get the data? Day et al., 2007 Celsius project Only Affymetrix sources shown Types of data • Single-color arrays (mainly Affymetrix) – Data reported as expression values – Raw values or log2-transformed values (RMA; GCRMA) • Two-color arrays – Data reported as expression ratios – Raw ratios or log2-transformed ratios Clustering with Cluster 3.0 • Based on original clustering program by Michael Eisen • Code updated by Michiel de Hoon • Runs on Windows, Mac, and Linux • Free from http://bonsai.ims.u-tokyo.ac.jp/~mdehoon/software/cluster/software.htm • Hierarchical, k-means, SOMs • Other option for large datasets: – XCluster, a command-line tool by Gavin Sherlock Getting Cluster 3.0 Cluster data import • Minimal matrix