A Fully Bayesian Model to Cluster Gene Expression Profiles

Clustering methods are popular screening tools for microarray data in order to identify subgroups of genes that share common regulatory elements, a common function or a common cellular origin. But the most popular clustering algorithms, e.g. K-means, require a priori determination of the number of clusters. Results strongly depend on this choice. Additionally, microarray data are inherently noisy and many measurements are missing, which results in the loss of a great amount of information with most earlier methods. Therefore, we propose a in which the and are treated as random variables that can be using the Reversible Jump Markov Chain Montecarlo (RJMCMC) simulation scheme [1]. probabilistic model number of clusters missing values estimated from the available dat