Comparisons of clustering algorithms for groupping genes based of expression profiles
by
Susmita Datta
Georgia State University

With the advent of microarray chip technology, large data sets are now emerging containing the simultaneous expression levels of thousands of genes at various time points during a biological process. Biologists are attempting to group genes based on the temporal pattern of their expression levels. In this talk we have selected a number of clustering algorithms (of various flavors!) which can be used to group genes based on their expression profiles. We study their performance on a well known publicly available microarray dataset on sporulation of budding yeast, as well as a simulated dataset. Among other things, we formulate three reasonable validation strategies that can be used for any microarray data when temporal observations or replications are present. We evaluate each of these clustering methods with these validation measures to see which is a good choice for a given dataset.

Date received: August 22, 2002