Reconstruction of a genetic network from gene perturbation data

We propose a new method for reconstructing a gene regulatory network from a gene expression profile. In this method we directly use the gene expression level as a weight of weighted network between genes. Each gene expression level is checked by means of a p-value test for each experiment to gain the confidence of the genetic network. We consider two different kinds of links to represent up and down regulation of the genes. Up (Down) regulation links are assigned between two genes when the gene expression level of one gene is increasing (decreasing) due to the other gene. To decipher the most prudent genetic network without degeneracy, we adapt the minimum spanning tree (MST) technique on weighted directional networks. The algorithm removes surplus links which originate from indirect effects of perturbation and results in the most concise directed network, containing two different kinds of links: positive-weight (activation) and negative-weight (repression) links. Applying the algorithm to a S. cerevisiae gene expression profile, we obtain the topology of the genetic network and find that the genetic network of yeast has a very inhomogeneous structure following the power-law degree distribution. We compare our result with a part of the genetic network of yeast known from the literature and databases.