## Applying dynamic bayesian networks to perturbed gene expression data (2006)

Venue: | BMC bioinformatics |

Citations: | 8 - 0 self |

### BibTeX

@ARTICLE{Dojer06applyingdynamic,

author = {Norbert Dojer and Anna Gambin and Jerzy Tiuryn},

title = {Applying dynamic bayesian networks to perturbed gene expression data},

journal = {BMC bioinformatics},

year = {2006},

volume = {7},

pages = {249}

}

### OpenURL

### Abstract

Abstract Motivation: A central goal of molecular biology is to understand the regulatory mechanisms of gene transcription and protein synthesis. Because of their solid basis in statistics, allowing to deal with the stochastic aspects of gene expressions and noisy measurements in a natural way, Bayesian networks appear attractive in the field of inferring gene interactions structure from microarray experiments data. However, the basic formalism has some disadvantages, e.g. it is sometimes hard to distinguish between the origin and the object of an interaction. Two kinds of microarray experiments yield data particularly rich in information regarding the direction of interactions: time series and perturbation experiments. In order to correctly handle them, the basic formalism must be modified. For example, dynamic Bayesian networks apply to time series microarray data. Results: We extend the framework of dynamic Bayesian networks in order to handle perturbations. A new discretization method, specialized for datasets from time series perturbations experiments, is also introduced. We compare networks inferred from realistic simulations data by our method and by dynamic Bayesian networks learning techniques. We conclude that application of our method substantially improves inferring. 1 Introduction As most genetic regulatory systems involve many components connected through complex networks of interactions, formal methods and computer tools for modeling and simulating are needed. Therefore, various formalisms were proposed to describe genetic regulatory systems, including Boolean networks and their generalizations, ordinary and partial differential equations, stochastic equations and Bayesian networks (see [4] for a review). While differential and stochastic equations describe the biophysical processes at a very refined level of detail and prove useful in simulations of well studied systems, Bayesian networks appear attractive in the field of inferring the regulatory network structure from gene expression data. The reason is that their learning techniques have solid basis in statistics, allowing to deal with the stochastic aspects of gene expressions and noisy measurements in a natural way.

### Citations

810 | Using Bayesian Networks to Analyze Expression Data
- Friedman, Linial, et al.
- 2000
(Show Context)
Citation Context ...equently, one has to choose between restricting to small gene networks (a relatively quick algorithm was given in [13]) and inferring suboptimal networks by heuristic search methods (see for instance =-=[5]-=-). It should be also pointed out that the basic BN formalism has some major limitations. First, several networks with the same undirected graph structure but different directions of some edges may rep... |

475 | Modeling and simulation of genetic regulatory systems: a literature review
- Jong
- 2002
(Show Context)
Citation Context ...ere proposed to describe genetic regulatory systems, including Boolean networks and their generalizations, ordinary and partial differential equations, stochastic equations and Bayesian networks (see =-=[4]-=- for a review). While differential and stochastic equations describe the biophysical processes at a very refined level of detail and prove useful in simulations of well studied systems, Bayesian netwo... |

251 |
Construction of a genetic toggle switch in Escherichia coli
- Gardner, Cantor, et al.
- 2000
(Show Context)
Citation Context ... protein and dimer X, respectively, finally X \DeltasY stands for a transcription factor bound to a promoter. The system is composed of structures reported in the biological literature (see [1], [2], =-=[7]-=-), i.e. a hysteretic oscillator, a genetic switch, cascades and a ligand binding mechanism that influences transcription (during the simulation, the ligand is injected for a short time). The whole net... |

229 | Learning the structure of dynamic probabilistic networks
- Friedman, Murphy, et al.
- 1998
(Show Context)
Citation Context ...relaxed. Given a set of time series of expression data, the learning techniques adapted from BNs allow one to infer dynamic networks that well match the temporal evolution recorded in the series (see =-=[6]-=-, [11], [14], [12], [15], [8], [9], [18]). A special treatment is required for experiments, in which expression of some genes were perturbed (e.g. knockout experiments). Since perturbations change the... |

168 | Modeling gene expression data using dynamic bayesian networks
- Murphy, Mian
(Show Context)
Citation Context ...ed. Given a set of time series of expression data, the learning techniques adapted from BNs allow one to infer dynamic networks that well match the temporal evolution recorded in the series (see [6], =-=[11]-=-, [14], [12], [15], [8], [9], [18]). A special treatment is required for experiments, in which expression of some genes were perturbed (e.g. knockout experiments). Since perturbations change the struc... |

168 | Inferring subnetworks from perturbed expression profiles
- Pe’er
- 2001
(Show Context)
Citation Context ...ven a set of time series of expression data, the learning techniques adapted from BNs allow one to infer dynamic networks that well match the temporal evolution recorded in the series (see [6], [11], =-=[14]-=-, [12], [15], [8], [9], [18]). A special treatment is required for experiments, in which expression of some genes were perturbed (e.g. knockout experiments). Since perturbations change the structure o... |

162 | Learning Bayesian network is NP-complete
- Chickering
- 1996
(Show Context)
Citation Context ...tions describe these interactions. Given a set of gene expression data, the learning techniques for Bayesian networks allow one to infer networks that match this set well. However, as it was shown in =-=[3]-=-, the problem of finding an optimal network is NP-hard. Consequently, one has to choose between restricting to small gene networks (a relatively quick algorithm was given in [13]) and inferring subopt... |

119 | Sensitivity and specificity of inferring genetic regulatory interactions from microarray experiments with dynamic Bayesian networks
- Husmeier
(Show Context)
Citation Context ...series of expression data, the learning techniques adapted from BNs allow one to infer dynamic networks that well match the temporal evolution recorded in the series (see [6], [11], [14], [12], [15], =-=[8]-=-, [9], [18]). A special treatment is required for experiments, in which expression of some genes were perturbed (e.g. knockout experiments). Since perturbations change the structure of interactions (r... |

56 | Dynamic Bayesian network and nonparametric regression for nonlinear modeling of gene networks from time series gene expression data - Kim - 2004 |

55 |
Gene networks inference using dynamic Bayesian networks
- PERRIN, RALAIVOLA, et al.
- 2003
(Show Context)
Citation Context ... time series of expression data, the learning techniques adapted from BNs allow one to infer dynamic networks that well match the temporal evolution recorded in the series (see [6], [11], [14], [12], =-=[15]-=-, [8], [9], [18]). A special treatment is required for experiments, in which expression of some genes were perturbed (e.g. knockout experiments). Since perturbations change the structure of interactio... |

53 | Modelling regulatory pathways in E. coli from time series expression profiles
- Ong, Glasner, et al.
- 2002
(Show Context)
Citation Context ...set of time series of expression data, the learning techniques adapted from BNs allow one to infer dynamic networks that well match the temporal evolution recorded in the series (see [6], [11], [14], =-=[12]-=-, [15], [8], [9], [18]). A special treatment is required for experiments, in which expression of some genes were perturbed (e.g. knockout experiments). Since perturbations change the structure of inte... |

45 |
How to make a biological switch
- JL, FR
- 2000
(Show Context)
Citation Context ...mRNA, protein and dimer X, respectively, finally X \DeltasY stands for a transcription factor bound to a promoter. The system is composed of structures reported in the biological literature (see [1], =-=[2]-=-, [7]), i.e. a hysteretic oscillator, a genetic switch, cascades and a ligand binding mechanism that influences transcription (during the simulation, the ligand is injected for a short time). The whol... |

43 | Discovery of causal relationships in a gene-regulation pathway from a mixture of experimental and observational DNA microarray data
- Yoo, Thorsson, et al.
- 2002
(Show Context)
Citation Context ...ns (regulation of affected genes is excluded), the learning techniques have to use data selectively. Inferring networks from perturbed expression profiles by means of BNs was investigated in [14] and =-=[16]-=-. In this report we extend the framework of DBN in order to handle perturbations and introduce a discretization method specialized for datasets from time series perturbations experiments. We compare n... |

38 |
A new Dynamic Bayesian Network (DBN) approach for identifying gene regulatory networks from time course microarray data
- Zou, Conzen
- 2005
(Show Context)
Citation Context ...expression data, the learning techniques adapted from BNs allow one to infer dynamic networks that well match the temporal evolution recorded in the series (see [6], [11], [14], [12], [15], [8], [9], =-=[18]-=-). A special treatment is required for experiments, in which expression of some genes were perturbed (e.g. knockout experiments). Since perturbations change the structure of interactions (regulation o... |

36 |
Bayes Net Toolbox for
- Murphy
(Show Context)
Citation Context ...The gene regulatory network from [17]. Rectangles denote promoters, zigzags stand for mRNAs and circles indicate proteins, dimers and a ligand (Q) Carlo method provided in Bayes Net Toolbox by Murphy =-=[10]-=-, is shown on Fig. 2(b). In the present study we introduce perturbations to the model. It is done by replacing the differential equation regarding the mRNA of a knocked out gene by the following equat... |

28 | Finding optimal models for small gene networks
- Ott, Imoto, et al.
- 2004
(Show Context)
Citation Context ...er, as it was shown in [3], the problem of finding an optimal network is NP-hard. Consequently, one has to choose between restricting to small gene networks (a relatively quick algorithm was given in =-=[13]-=-) and inferring suboptimal networks by heuristic search methods (see for instance [5]). It should be also pointed out that the basic BN formalism has some major limitations. First, several networks wi... |

20 | Simulation studies for the identification of genetic networks from cDNA array and regulatory activity data
- Zak, Doyle, et al.
- 2001
(Show Context)
Citation Context ...between the above two extremes is to apply the learning procedure to data generated by a system of ordinary differentional equations. In the present study we generate data using the model proposed in =-=[17]-=-. The model consists of 54 species of molecules, representing 10 genes with their transcription factors, promoters, mRNAs, proteins and protein dimers, connected through 97 elementary reactions, inclu... |

12 |
Circadian clocks limited by noise. Nature
- Barkai, Leibler
- 2000
(Show Context)
Citation Context ...ter, mRNA, protein and dimer X, respectively, finally X \DeltasY stands for a transcription factor bound to a promoter. The system is composed of structures reported in the biological literature (see =-=[1]-=-, [2], [7]), i.e. a hysteretic oscillator, a genetic switch, cascades and a ligand binding mechanism that influences transcription (during the simulation, the ligand is injected for a short time). The... |