Biological networks inherently have uncertain topologies. This arises from many factors. For instance, interactions\nbetween molecules may or may not take place under varying conditions. Genetic or epigenetic mutations may also\nalter biological processes like transcription or translation. This uncertainty is often modeled by associating each\ninteraction with a probability value. Studying biological networks under this probabilistic model has already been\nshown to yield accurate and insightful analysis of interaction data. However, the problem of assigning accurate\nprobability values to interactions remains unresolved. In this paper, we present a novel method for computing\ninteraction probabilities in signaling networks based on transcription levels of genes. The transcription levels define\nthe signal reachability probability between membrane receptors and transcription factors. Our method computes the\ninteraction probabilities that minimize the gap between the observed and the computed signal reachability\nprobabilities. We evaluate our method on four signaling networks from the Kyoto Encyclopedia of Genes and\nGenomes (KEGG). For each network, we compute its edge probabilities using the gene expression profiles for seven\nmajor leukemia subtypes. We use these values to analyze how the stress induced by different leukemia subtypes\naffects signaling interactions.
Loading....