Limits...
GBNet: deciphering regulatory rules in the co-regulated genes using a Gibbs sampler enhanced Bayesian network approach.

Shen L, Liu J, Wang W - BMC Bioinformatics (2008)

Bottom Line: Most of the rules learned by GBNet on YY1 and co-factors were supported by literature.In addition, a spacing constraint between YY1 and E2F was also supported by independent TF binding experiments.We thus conclude that GBNet is a useful tool for deciphering the "grammar" of transcriptional regulation.

View Article: PubMed Central - HTML - PubMed

Affiliation: Department of Chemistry and Biochemistry, University of California, San Diego, California, USA. shen@ucsd.edu

ABSTRACT

Background: Combinatorial regulation of transcription factors (TFs) is important in determining the complex gene expression patterns particularly in higher organisms. Deciphering regulatory rules between cooperative TFs is a critical step towards understanding the mechanisms of combinatorial regulation.

Results: We present here a Bayesian network approach called GBNet to search for DNA motifs that may be cooperative in transcriptional regulation and the sequence constraints that these motifs may satisfy. We showed that GBNet outperformed the other available methods in the simulated and the yeast data. We also demonstrated the usefulness of GBNet on learning regulatory rules between YY1, a human TF, and its co-factors. Most of the rules learned by GBNet on YY1 and co-factors were supported by literature. In addition, a spacing constraint between YY1 and E2F was also supported by independent TF binding experiments.

Conclusion: We thus conclude that GBNet is a useful tool for deciphering the "grammar" of transcriptional regulation.

Show MeSH
An example of the Bayesian network learning procedure in BBNet and GBNet. The sequences were taken from the fourth yeast cluster in [16]. The magenta line represents the landscape of the Bayesian score (absolute value). The learning steps involving motifs other than PAC and RRPE were omitted for the illustration purpose. The parent nodes of the regulator rules learned in the three key steps are shown on the right.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC2571992&req=5

Figure 2: An example of the Bayesian network learning procedure in BBNet and GBNet. The sequences were taken from the fourth yeast cluster in [16]. The magenta line represents the landscape of the Bayesian score (absolute value). The learning steps involving motifs other than PAC and RRPE were omitted for the illustration purpose. The parent nodes of the regulator rules learned in the three key steps are shown on the right.

Mentions: To illustrate why the GBNet could but BBNet could not find the two rules, we examined each step of the Bayesian network structure learning (Fig. 2). When the searching reached a local optimum (state 1 in Fig. 2) with a Bayesian score of -101.3, the network contained two parent nodes (Fig. 2): "distance to ATG of PAC" and "presence of RRPE". If the "distance to ATG of RRPE" node was added, the Bayesian score would decrease. Therefore, the greedy search in BBNet stopped and did not add this rule. The searching was thus trapped in the local optimum. In contrast, a Metropolis jump was tried in GBNet with an accepting probability calculated based on the difference of the Bayesian scores before and after the jump (see Methods): the closer the two Bayesian scores, the more likely a jump got accepted. To further enhance the sampling power, simulated annealing was also employed in GBNet and multiple iterations were executed until the model was converged at a specific temperature. As a result of this searching strategy, the "distance to ATG of RRPE" rule was added by GBNet even though the Bayesian score became worse (state 2 in Fig. 2): -103.93 versus -101.3. Next, the Bayesian score was improved to -96.26 by removing the "presence of RRPE" node (state 3 in Fig. 2). The two correct rules were thus found and being kept to the end of the searching. This example illustrates the advantages of the searching strategy implemented in GBNet to avoid being trapped in the local optimum compared with the greedy search algorithm in BBNet.


GBNet: deciphering regulatory rules in the co-regulated genes using a Gibbs sampler enhanced Bayesian network approach.

Shen L, Liu J, Wang W - BMC Bioinformatics (2008)

An example of the Bayesian network learning procedure in BBNet and GBNet. The sequences were taken from the fourth yeast cluster in [16]. The magenta line represents the landscape of the Bayesian score (absolute value). The learning steps involving motifs other than PAC and RRPE were omitted for the illustration purpose. The parent nodes of the regulator rules learned in the three key steps are shown on the right.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC2571992&req=5

Figure 2: An example of the Bayesian network learning procedure in BBNet and GBNet. The sequences were taken from the fourth yeast cluster in [16]. The magenta line represents the landscape of the Bayesian score (absolute value). The learning steps involving motifs other than PAC and RRPE were omitted for the illustration purpose. The parent nodes of the regulator rules learned in the three key steps are shown on the right.
Mentions: To illustrate why the GBNet could but BBNet could not find the two rules, we examined each step of the Bayesian network structure learning (Fig. 2). When the searching reached a local optimum (state 1 in Fig. 2) with a Bayesian score of -101.3, the network contained two parent nodes (Fig. 2): "distance to ATG of PAC" and "presence of RRPE". If the "distance to ATG of RRPE" node was added, the Bayesian score would decrease. Therefore, the greedy search in BBNet stopped and did not add this rule. The searching was thus trapped in the local optimum. In contrast, a Metropolis jump was tried in GBNet with an accepting probability calculated based on the difference of the Bayesian scores before and after the jump (see Methods): the closer the two Bayesian scores, the more likely a jump got accepted. To further enhance the sampling power, simulated annealing was also employed in GBNet and multiple iterations were executed until the model was converged at a specific temperature. As a result of this searching strategy, the "distance to ATG of RRPE" rule was added by GBNet even though the Bayesian score became worse (state 2 in Fig. 2): -103.93 versus -101.3. Next, the Bayesian score was improved to -96.26 by removing the "presence of RRPE" node (state 3 in Fig. 2). The two correct rules were thus found and being kept to the end of the searching. This example illustrates the advantages of the searching strategy implemented in GBNet to avoid being trapped in the local optimum compared with the greedy search algorithm in BBNet.

Bottom Line: Most of the rules learned by GBNet on YY1 and co-factors were supported by literature.In addition, a spacing constraint between YY1 and E2F was also supported by independent TF binding experiments.We thus conclude that GBNet is a useful tool for deciphering the "grammar" of transcriptional regulation.

View Article: PubMed Central - HTML - PubMed

Affiliation: Department of Chemistry and Biochemistry, University of California, San Diego, California, USA. shen@ucsd.edu

ABSTRACT

Background: Combinatorial regulation of transcription factors (TFs) is important in determining the complex gene expression patterns particularly in higher organisms. Deciphering regulatory rules between cooperative TFs is a critical step towards understanding the mechanisms of combinatorial regulation.

Results: We present here a Bayesian network approach called GBNet to search for DNA motifs that may be cooperative in transcriptional regulation and the sequence constraints that these motifs may satisfy. We showed that GBNet outperformed the other available methods in the simulated and the yeast data. We also demonstrated the usefulness of GBNet on learning regulatory rules between YY1, a human TF, and its co-factors. Most of the rules learned by GBNet on YY1 and co-factors were supported by literature. In addition, a spacing constraint between YY1 and E2F was also supported by independent TF binding experiments.

Conclusion: We thus conclude that GBNet is a useful tool for deciphering the "grammar" of transcriptional regulation.

Show MeSH