Studying the Conditions for Learning Dynamic Bayesian Networks to Discover Genetic Regulatory Networks

Handling Data Sparseness in Gene Network Reconstruction

“…The continuous Bayesian network using ARIA will be compared with a discrete BN in its capability of recovering the original structure based solely on the data generated by the true network. The structure learning search heuristic used here (which will be the same for both BNs models) is the greedy hill climbing strategy, since it is reported in [25] to have the best performance when compared to other algorithms. In order to discretize the variables, we employed the method proposed in [12], which uses k-means clustering to detect three discrete expression levels.…”

“…In order to discretize the variables, we employed the method proposed in [12], which uses k-means clustering to detect three discrete expression levels. This is a very sophisticated discretization method and has also been used elsewhere [25].…”

“…Experiments with real data would be restricted to subjective analysis and conclusions [8,25]. As outlined in [26], given the perfect knowledge of the true underlying structure and the possibility of precise control of the quantity and quality of data, synthetic networks are still the best way to carry out an objective comparison between different methods.…”

Reverse engineering and verification of gene networks: Principles, assumptions, and limitations of present methods and future perspectives

“…The network interactions are drawn from the genes expressed at an earlier time-point to those at the later time. By using the time-series data, DBN might be able to identify the feedback loops (Kim, 2007;Perrin et al, 2003;van Berlo, 2003;Zou and Conzen, 2005).…”

Inferring the skeleton cell cycle regulatory network of malaria parasite using comparative genomic and variational Bayesian approaches

“…Various linear and nonlinear models for continuous and discrete gene expression levels have been studied for gene regulation. Among the models that consider discrete gene expression levels, the discrete Bayesian networks (Gat-Viks et al 2005;Pe'er et al 2001;Perrin et al 2003;van Berlo et al 2003), the Probabilistic Boolean networks (Dougherty et al 2005;Shmulevich et al 2002) and the Gaussian Graphical models (Schafer and Strimmer 2005;Wu et al 2003) deserve attention. Their main advantage is low computational complexity and an ability to capture non-linearity in regulatory functions.…”