Advances in Statistical Modeling of High Dimensional Data:
Variable selection, and Challenges in Image Analysis

Bettina Knapp, Bioquant Heidelberg

Reverse Engineering of Signaling Pathways from RNAi Data

The inference of signal transduction and genetic regulatory networks is a major goal in systems biology. Systematic screenings of RNA interference (RNAi) offer the possibility to identify genes related with a particular phenotype or cellular pathway of interest (Fire, 1998). The temporal and spatial placement of these genes in the respective cellular pathway remains a challenging problem (Moffat and Sabatini, 2006). While Sacher et al. (2008) cluster phenotypes, Markowetz et al. (2007) use the nested structure of effects of different knockdowns to solve this problem. We propose a stochastic model with Boolean networks for pathway inference where the activation probabilities for each gene are described by sigmoid functions. A Markov chain Monte Carlo approach is used to infer model topology and model parameters simultaneously, by sampling from the posterior distribution over model parameters given the knockdown data in a Bayesian setting. We compute the exact transition probabilities between different network states using the effect of single- or combinatorial knockdowns. Incomplete observations are integrated out via marginalization over unobserved nodes. To address the problem of under determined model parameters we use a prior distribution on the model parameters. We then approximate the likelihood allowing to sample from the posterior distribution without explicit evaluation of the likelihood whereas the sampling from the posterior for networks with larger number of nodes is permitted. We evaluated our method on a small artificial network with five nodes and we present results from the inference of the Jak/Stat signal transduction pathway in a hepatoma cell line given the knockdown data of 11 genes of the core Jak/Stat pathway.

[Back to the Schedule]