| |
1 Institute of Animal Breeding and Genetics, Veterinaermedizinische
Universitaet Wien, 1210 Vienna, Austria
2 Institute for Genomics and Bioinformatics, Graz University
of Technology, 8010 Graz, Austria
3 Faculty of Life Sciences, University of Manchester, M60 1QD
Manchester, UK
4 Institute for Informatic, University of Rostock, 18051 Rostock,
Germany
* these authors contributed equally
|
| | Cell cycle, organ development, and cellular differentiation involve
regular cascades of changes in gene expression. With cDNA or
oligonucleotide chips, these changes can be simultaneously monitored
for most genes in a genome. After proper normalization of the data,
genes are often classified into co-expressed classes (clusters) to
identify subgroups of genes that share common regulatory elements, a
common function, or a common cellular origin.
We propose a fully probabilistic Bayesian model to cluster gene
expression profiles. The number of classes does not need to be
specified in advance, instead it is adjusted dynamically using a
Reversible Jump Markov Chain Monte Carlo (RJMCMC) algorithm. In
addition, the imputation of missing values is integrated into the
model. Simulated data was used to assess the performance of the algorithm.
Specificity was very high while sensitivity was around 50%, outperforming the results from the traditional k-means algorithm.
Clusters from data sets with and without missing values showed a great simmilarity.
The method is specially useful in order to determine genes likely to be involved in the same biological process than a given
one or to identify genes that exhibit a pre-determined profile relevant to the process under study.
|
the clusters
with and without missing values.
