Limits...
Supervised inference of gene-regulatory networks.

To CC, Vohradsky J - BMC Bioinformatics (2008)

Bottom Line: The results were compared with independent data sources.Results show that our algorithm gives, in most cases, results identical with the independent experiments, when compared with the YEASTRACT database.In several cases our algorithm gives predictions of novel interactions which have not been reported.

View Article: PubMed Central - HTML - PubMed

Affiliation: Laboratory of Bioinformatics, Institute of Microbiology ASCR, Prague, Czech Republic. cuongto@biomed.cas.cz

ABSTRACT

Background: Inference of protein interaction networks from various sources of data has become an important topic of both systems and computational biology. Here we present a supervised approach to identification of gene expression regulatory networks.

Results: The method is based on a kernel approach accompanied with genetic programming. As a data source, the method utilizes gene expression time series for prediction of interactions among regulatory proteins and their target genes. The performance of the method was verified using Saccharomyces cerevisiae cell cycle and DNA/RNA/protein biosynthesis gene expression data. The results were compared with independent data sources. Finally, a prediction of novel interactions within yeast gene expression circuits has been performed.

Conclusion: Results show that our algorithm gives, in most cases, results identical with the independent experiments, when compared with the YEASTRACT database. In several cases our algorithm gives predictions of novel interactions which have not been reported.

Show MeSH
Interaction networks of genes adopted from the work of Lee et al. [6] with sub-networks (bold) used as a training set. Shaded nodes represent genes for which the regulatory   interactions were predicted using the algorithm presented here.   A – cell cycle network, B – DNA/RNA/protein synthesis network.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC2266705&req=5

Figure 2: Interaction networks of genes adopted from the work of Lee et al. [6] with sub-networks (bold) used as a training set. Shaded nodes represent genes for which the regulatory interactions were predicted using the algorithm presented here. A – cell cycle network, B – DNA/RNA/protein synthesis network.

Mentions: Two protein networks, namely cell cycle and DNA/RNA/Protein biosynthesis identified by Lee et al., served here as a template for comparison with the results of the algorithm we have now presented. Parts of the networks were used as training sets (Figure 2) and the remaining interactions were inferred using the trained algorithm and the genes presented in the work of Lee et al. Such arrangement simulates a situation where only a very limited part of a network is known. In reality, such sub-network can be inferred either from measurements or from a literature surveys. Here we identify the rest of the network using the presented algorithm. In this test example, the complete network is known a priori (we consider the Lee's et al. network as complete for comparison purposes). The prior knowledge allows us to assess the performance of the algorithm by comparison of predicted interactions and interactions inferred from the independent source, the work of Lee et al. For this reason the same set of genes as in the work of Lee at al. was used (the full set of genes is depicted in Figure 7 (see Additional file 1)). The trained algorithm was applied to the expression profiles of these selected genes. The networks inferred by Lee et al. and the selected training sub-networks are plotted in Figure 2. The results, i.e. predicted interactions, are listed in Table 2 and Figure 7 (see Additional file 1). For the independent verification of the results of our algorithm and the experimental results of Lee et al., information about the documented and potential interactions among yeast genes and gene products from the YEASTRACT database was used. The YEASTRACT (Yeast Search for Transcriptional Regulators And Consensus Tracking) represents one of the most comprehensive data sources about regulatory interactions in yeast. It is a curated repository which, in the time of publication of this paper, comprised more than 12500 regulatory associations between transcription factors and target genes in Saccharomyces cerevisiae, based on more than 900 bibliographic references. It also included the description of 269 specific DNA binding sites for more than a hundred characterized transcription factors.


Supervised inference of gene-regulatory networks.

To CC, Vohradsky J - BMC Bioinformatics (2008)

Interaction networks of genes adopted from the work of Lee et al. [6] with sub-networks (bold) used as a training set. Shaded nodes represent genes for which the regulatory   interactions were predicted using the algorithm presented here.   A – cell cycle network, B – DNA/RNA/protein synthesis network.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC2266705&req=5

Figure 2: Interaction networks of genes adopted from the work of Lee et al. [6] with sub-networks (bold) used as a training set. Shaded nodes represent genes for which the regulatory interactions were predicted using the algorithm presented here. A – cell cycle network, B – DNA/RNA/protein synthesis network.
Mentions: Two protein networks, namely cell cycle and DNA/RNA/Protein biosynthesis identified by Lee et al., served here as a template for comparison with the results of the algorithm we have now presented. Parts of the networks were used as training sets (Figure 2) and the remaining interactions were inferred using the trained algorithm and the genes presented in the work of Lee et al. Such arrangement simulates a situation where only a very limited part of a network is known. In reality, such sub-network can be inferred either from measurements or from a literature surveys. Here we identify the rest of the network using the presented algorithm. In this test example, the complete network is known a priori (we consider the Lee's et al. network as complete for comparison purposes). The prior knowledge allows us to assess the performance of the algorithm by comparison of predicted interactions and interactions inferred from the independent source, the work of Lee et al. For this reason the same set of genes as in the work of Lee at al. was used (the full set of genes is depicted in Figure 7 (see Additional file 1)). The trained algorithm was applied to the expression profiles of these selected genes. The networks inferred by Lee et al. and the selected training sub-networks are plotted in Figure 2. The results, i.e. predicted interactions, are listed in Table 2 and Figure 7 (see Additional file 1). For the independent verification of the results of our algorithm and the experimental results of Lee et al., information about the documented and potential interactions among yeast genes and gene products from the YEASTRACT database was used. The YEASTRACT (Yeast Search for Transcriptional Regulators And Consensus Tracking) represents one of the most comprehensive data sources about regulatory interactions in yeast. It is a curated repository which, in the time of publication of this paper, comprised more than 12500 regulatory associations between transcription factors and target genes in Saccharomyces cerevisiae, based on more than 900 bibliographic references. It also included the description of 269 specific DNA binding sites for more than a hundred characterized transcription factors.

Bottom Line: The results were compared with independent data sources.Results show that our algorithm gives, in most cases, results identical with the independent experiments, when compared with the YEASTRACT database.In several cases our algorithm gives predictions of novel interactions which have not been reported.

View Article: PubMed Central - HTML - PubMed

Affiliation: Laboratory of Bioinformatics, Institute of Microbiology ASCR, Prague, Czech Republic. cuongto@biomed.cas.cz

ABSTRACT

Background: Inference of protein interaction networks from various sources of data has become an important topic of both systems and computational biology. Here we present a supervised approach to identification of gene expression regulatory networks.

Results: The method is based on a kernel approach accompanied with genetic programming. As a data source, the method utilizes gene expression time series for prediction of interactions among regulatory proteins and their target genes. The performance of the method was verified using Saccharomyces cerevisiae cell cycle and DNA/RNA/protein biosynthesis gene expression data. The results were compared with independent data sources. Finally, a prediction of novel interactions within yeast gene expression circuits has been performed.

Conclusion: Results show that our algorithm gives, in most cases, results identical with the independent experiments, when compared with the YEASTRACT database. In several cases our algorithm gives predictions of novel interactions which have not been reported.

Show MeSH