Learning the structure of gene regulatory networks from time series gene expression data
© Li et al. licensee BioMed Central Ltd 2011
Published: 23 December 2011
Dynamic Bayesian Network (DBN) is an approach widely used for reconstruction of gene regulatory networks from time-series microarray data. Its performance in network reconstruction depends on a structure learning algorithm. REVEAL (REVerse Engineering ALgorithm) is one of the algorithms implemented for learning DBN structure and used to reconstruct gene regulatory networks (GRN). However, the two-stage temporal Bayes network (2TBN) structure of DBN that specifies correlation between time slices cannot be obtained by score metrics used in REVEAL.
In this paper, we study a more sophisticated score function for DBN first proposed by Nir Friedman for stationary DBNs structure learning of both initial and transition networks but has not yet been used for reconstruction of GRNs. We implemented Friedman's Bayesian Information Criterion (BIC) score function, modified K2 algorithm to learn Dynamic Bayesian Network structure with the score function and tested the performance of the algorithm for GRN reconstruction with synthetic time series gene expression data generated by GeneNetWeaver and real yeast benchmark experiment data.
We implemented an algorithm for DBN structure learning with Friedman's score function, tested it on reconstruction of both synthetic networks and real yeast networks and compared it with REVEAL in the absence or presence of preprocessed network generated by Zou&Conzen's algorithm. By introducing a stationary correlation between two consecutive time slices, Friedman's score function showed a higher precision and recall than the naive REVEAL algorithm.
Friedman's score metrics for DBN can be used to reconstruct transition networks and has a great potential to improve the accuracy of gene regulatory network structure prediction with time series gene expression datasets.
High-content technologies such as DNA microarrays can provide a system-scale overview of how genes interact with each other in a network context. This network is called a gene regulatory network (GRN) and can be defined as a mixed graph over a set of nodes (corresponding to genes or gene activities) with directed or undirected edges (representing causal interactions or associations between gene activities) . Various mathematical methods and computational approaches have been proposed to reconstruct GRNs, including Boolean networks , information theory [3, 4], differential equations  and Bayesian networks [6–8]. GRN reconstruction faces huge intrinsic challenges on both experimental and theoretical fronts, because the inputs and outputs of the molecular processes are unclear and the underlying principles are unknown or too complex. In the previous work, we compared two important computational approaches, Dynamic Bayesian networks (DBNs) and Probabilistic Boolean networks for reconstructing GRNs using a time-series dataset from the Drosophila Interaction Database, and found that DBN outperforms PBN . In this paper, we emphasize the DBN approach.
Dynamic Bayesian networks (DBNs) are belief networks that represent the stochastic process of a set of random variables over time. The hidden Markov model (HMM) and the Kalman filter can be considered as the simplest DBNs. However, Kalman filters can only handle unimodal posterior distributions and linear models, whereas parameterization of HMM grows exponentially with the number of state variables . Several algorithms have been developed to learn structure for belief networks from both complete [6, 10–12] (without missing values) and incomplete [13, 14] (with missing values) datasets. Structure Expectation-Maximization (SEM) has been developed for learning Probabilistic network structure from data with hidden variables and missing values . A structure learning algorithm has also been developed for high-order and non-stationary dynamic probabilistic models .
A commonly used structure learning algorithm is based on REVEAL (REVerse Engineering ALgorithm) [6, 12] which learns the optimal set of parents for each node of a network independently, based on an information theoretic concept of mutual information analysis. However, the two-stage temporal Bayes network (2TBN) cannot be well recovered by application of REVEAL. In this work, we implemented a more sophisticated algorithm, proposed by Friedman , to learn the structure of both initial networks and transition networks, which specified a stationary correlation between two consecutive time periods. Compared with Murphy's algorithm, it improves performance in two ways. First, in score function, it considers time lags that may happen in biological processes. Second, it fetches the relationship which gains the maximum score function in the same time period or in the two consecutive time periods. Thus, Friedman's DBN structure learning algorithm was used in our work and its performance in terms of reconstruction accuracy was also evaluated using synthetic gene expression datasets and a real yeast time-series benchmark dataset.
In the following sections, we first provide an introduction to DBN and existing DBN algorithms for reconstruction of GRNs. We then present an implementation of Friedman's DBN algorithm. Finally, we apply the algorithms to synthetic datasets and a real yeast benchmark dataset, and compare its performance to the commonly used Murphy's DBN algorithm [12, 16] based on REVEAL.
Dynamic Bayesian networks
Bayesian information criterion for DBN
where P(D|θG,G) is the marginal likelihood of the data given the network 〈G,θG〉 and P(θG|G) is our pior.
where [G] is the model dimension, or the number of independent parameters in G.
This approximation is called the Bayesian information criterion (BIC). N. Friedman, et al. deduce BIC for Dynamic Bayesian Network in his work, which is briefly described below.
It is assumed that the dataset D is composed of N seq complete observations. The first such sequence has length N i and specifies values for the variables x l ,...,x l [N l ], which means in different time slice the number of observations can be different. With such a dataset, we can learn B0 from N seq observations of initial slice, and learn B→ by transactions of transition slices.
where I(·;x l ) is an indicator function which equals 1 if the corresponding event occurs in sequence x l , and 0 otherwise.
Such decomposition implies that B0 is independent from B → , so we can give BIC score as BIC(G:D) = BIC0+BIC→
Learning network structure
Existing approaches for comparison
For convenience of performance analysis in the next section, we briefly describe Murphy and Zou's previous work here and present results in the next section. The widely used DBN implementation developed by Murphy and Mian (called Murphy's DBN hereafter) is based on REVEAL . Given an unknown structure with full observations, the algorithm learns the parent set for each node independently. There are 2n such sets, which can be arranged in a lattice for the permutation of factors. The problem is to find the highest score in the lattice. The approach taken by REVEAL is started from the bottom of the lattice, and evaluates the score at all points in the successive level until a point is found with a score of 1.0. Zou and Conzen  proposed a method to generate a preprocessed network for potential regulators by biological interpretation of time course microarray data. It assumes that the gene with earlier initial up-regulation is the potential regulator of those with later initial up-regulation. This preprocessed network is used to narrow down the search space for Murphy's DBN algorithm because it requires excessive time to find a permutation for each node even when imposing a maximum number of parents for the nodes if the network dimension is large.
Results and discussion
The Friedman's algorithm described in the method section was implemented based on Murphy's BNT tool box (Bayes Net Toolbox for Matlab). We tested four cases of DBN algorithms on reconstruction of synthetic networks. The four methods are: (1) Zou's preprocessed networks consisting of potential regulators by biological interpretation of time course microarray data (Zou&Conzen), (2) Murphy's DBN, implemented in conjunction with the preprocessed networks (Kevin Murphy + Zou&Conzen), (3) Friedman's algorithm (Nir Friedman), and (4) Friedman's algorithm combined with the preprocessed networks (Friedman + Zou&Conzen).
Precision (P) and recall (R) were used as the metrics for performance comparison. Here, R is defined as C e /(C e + M e ) and P is defined as C e /(C e + F e ), where C e denotes true positive edges that exist in both the true network and the predicted network, M e false negative edges that exist in the true network but not in the inferred network, and F e false positive edges that do not exist in the true network but do exist in the predicted network.
The synthetic datasets and network were generated using GeneNetWeaver from DREAM (Dialogue for Reverse Engineering Assessments and Methods) projects . We used sub-networks of different sizes (i.e., 10, 20, 50 and 100 genes) with randomly picked factors from a high-dimensional yeast GRN with 4441 nodes and 12873 edges. A model consisting of ordinary and stochastic differential equations and Gaussian noise model was used to generate synthetic gene expression data with a total of 21 time points and 10 replicates for each time slice.
Comparison of performance between different structure learning algorithms using synthetic dataset (Ce: Correctly infered edges; P: Precision; R: Recall)
Nir Friedman + Zou&Conzon
Kevin Murphy + Zou&Conzon
Real yeast benchmark dataset
Comparison of performance between different structure learning algorithms using yeast benchmark dataset (Ce: Correctly infered edges; P: Precision; R: Recall)
Kevin Murphy + Zou&Conzon
Probabilistic Boolean Network
In this study, we implemented Friedman's score metrics for DBNs by our algorithm, and applied the algorithm in reconstruction GRNs using both synthetic time series gene expression data and a real yeast benchmark dataset. The algorithm is able to capture the correlation between consecutive time slices in both score function and learning procedure, thus Friedman's score metrics gives a higher precision and recall than the naive REVEAL algorithm application in the absence or presence of preprocessed network generated by Zou&Conzen's algorithm. This also reflects that in real biological processes, time lag regulation might better describe the true regulation between genes. Also based on the testing results, the Friedman's score metrics we implemented has great potential in improving the accuracy of structure prediction for GRN reconstruction with complete synthetic time series data.
This work was supported by the US Army Corps of Engineers Environmental Quality Program under contract # W912HZ-08-2-0011and the NSF EPSCoR project "Modeling and Simulation of Complex Systems" (NSF #EPS-0903787). Permission was granted by the Chief of Engineers to publish this information.
- De la Fuente A: What are gene regulatory networks?. Handbook of Research on Computational Methodologies in Gene Regulatory Networks. Edited by: Das S, Caragea D, Welch SM, Hsu WH. 2010, Medical Information Science Reference, Hershey, PA, 1-27.View ArticleGoogle Scholar
- Lähdesmäki H, et al: On learning gene regulatory networks under the Boolean network model. Machine Learning. 2003, 52 (1,2): 147-167.View ArticleGoogle Scholar
- Faith JJ, et al: Large-scale mapping and validation of Escherichia coli transcriptional regulation from a compendium of expression profiles. PLoS Bio. 2007, 5 (1): e8-10.1371/journal.pbio.0050008.View ArticleGoogle Scholar
- Chaitankar Vijender, et al: A novel gene network inference algorithm using predictive minimum description length approach. BMC Syst Biol. 2010, 4 (Suppl 1): S7-10.1186/1752-0509-4-S1-S7.PubMed CentralView ArticlePubMedGoogle Scholar
- Chen T, et al: Modeling gene expression with differential equations. Pac Symp Biocomput. 1999, 4: 29-41.Google Scholar
- Liang S, et al: REVEAL, a general reverse engineering algorithm for inference of genetic network architectures. Pac Symp Biocomput. 1998, 3: 18-29.Google Scholar
- Imoto S, et al: Estimation of genetic networks and functional structures between genes by using Bayesian networks and nonparametric regression. Pac Symp Biocomput. 2002, 7: 175-186.Google Scholar
- Stolovitzky G, et al: Lessons from the DREAM2 Challenges. Ann N Y Acad Sci. 2009, 1158 (1): 159-195. 10.1111/j.1749-6632.2009.04497.x.View ArticlePubMedGoogle Scholar
- Li P, et al: Comparison of probabilistic Boolean network and dynamic Bayesian network approaches for inferring gene regulatory networks. BMC Bioinformatics. 2007, 8 (Suppl 8): S13-PubMed CentralView ArticlePubMedGoogle Scholar
- Friedman N, et al: Learning the structure of dynamic probabilistic networks. Proc UAI-98, Madison. 1998, Wisconsin: Morgan KaufmannGoogle Scholar
- Cooper GF, Herskovits E: A Bayesian method for the induction of probabilistic networks from data. Machine Learning. 1992, 9 (4): 309-347.Google Scholar
- Murphy KP, Mian S: Modeling Gene Expression Data Using Dynamic Bayesian Networks. Tech rep, MIT Artificial Intelligence Laboratory. 1999Google Scholar
- Friedman N: Learning belief networks in the presence of missing values and hidden variables. ICML-97. 1997Google Scholar
- Friedman N: The Bayesian Structural EM Algorithm. UAI-98. 1998Google Scholar
- Tucker A, Liu X: A Bayesian network approach to explaining time series with changing structure. Intell Data Anal. 2004, 8 (5):Google Scholar
- Murph K: Dynamic Bayesian Networks: Representation, Inference and Learning. PhD thesis, Computer Science Division, UC Berkeley. 2002Google Scholar
- Zou M, Conzen SD: A new dynamic Bayesian network (DBN) approach for identifying gene regulatory networks from time course microarray data. Bioinformatics. 2005, 21 (1): 71-79. 10.1093/bioinformatics/bth463.View ArticlePubMedGoogle Scholar
- Marbach D, et al: Generating realistic in silico gene networks for performance assessment of reverse engineering method. J Comput Biol. 2009, 16 (2): 229-239. 10.1089/cmb.2008.09TT.View ArticlePubMedGoogle Scholar
- Spellman , et al: Comprehensive identification of cell cycle-regulated genes of the yeast Saccharomyces cerevisiae by microarray hybridization. Mol Biol Cell. 1998, 9: 3273-3297.PubMed CentralView ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.