Florian Markowetz and Rainer Spang
Evaluating the Effect of Perturbations in Reconstructing Network Topologies
***************************************************************************

Many different Bayesian network models have been suggested to reconstruct gene
expression networks from microarray data. However, little attention has been
payed to the effects of small sample size and the stability of the solution.
We engage in a systematic investigation of these issues.

As a starting point for further research we introduce the kappa-network. It is
a small Bayesian network model (5 nodes with three states) in which a parameter
kappa controls the conditional probability distributions of the nodes. With data
sampled from this model, we evaluate the effects of different sample sizes and
of data being derived from active perturbations on the reconstruction of the
origninal network topology.