Limits...
A blind deconvolution approach to high-resolution mapping of transcription factor binding sites from ChIP-seq data.

Lun DS, Sherrid A, Weiner B, Sherman DR, Galagan JE - Genome Biol. (2009)

Bottom Line: We present CSDeconv, a computational method that determines locations of transcription factor binding from ChIP-seq data.CSDeconv differs from prior methods in that it uses a blind deconvolution approach that allows closely-spaced binding sites to be called accurately.We apply CSDeconv to novel ChIP-seq data for DosR binding in Mycobacterium tuberculosis and to existing data for GABP in humans and show that it can discriminate binding sites separated by as few as 40 bp.

View Article: PubMed Central - HTML - PubMed

Affiliation: Phenomics and Bioinformatics Research Centre, School of Mathematics and Statistics, and Australian Centre for Plant Functional Genomics, University of South Australia, Mawson Lakes Boulevard, Mawson Lakes, SA 5095, Australia. desmond.lun@unisa.edu.au

ABSTRACT
We present CSDeconv, a computational method that determines locations of transcription factor binding from ChIP-seq data. CSDeconv differs from prior methods in that it uses a blind deconvolution approach that allows closely-spaced binding sites to be called accurately. We apply CSDeconv to novel ChIP-seq data for DosR binding in Mycobacterium tuberculosis and to existing data for GABP in humans and show that it can discriminate binding sites separated by as few as 40 bp.

Show MeSH

Related in: MedlinePlus

Empirical FDR of CSDeconv. (a, b) The empirical FDR for enriched regions as a function of LLR threshold is shown for the (a) DosR and (b) GABP datasets. (c, d) With the LLR threshold fixed, the empirical FDR for binding sites as a function of the regularization factor α is shown for the (c) DosR dataset (LLR threshold at 18.75) and the (d) GABP dataset (LLR threshold at 38.5).
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC2812949&req=5

Figure 2: Empirical FDR of CSDeconv. (a, b) The empirical FDR for enriched regions as a function of LLR threshold is shown for the (a) DosR and (b) GABP datasets. (c, d) With the LLR threshold fixed, the empirical FDR for binding sites as a function of the regularization factor α is shown for the (c) DosR dataset (LLR threshold at 18.75) and the (d) GABP dataset (LLR threshold at 38.5).

Mentions: In Figures 2a and 2b, we show the empirical FDR for enriched regions as a function of the LLR threshold for the DosR and GABP datasets, respectively. We see that, owing to its lower coverage, larger LLR thresholds are required to achieve low FDRs in the GABP dataset. To ensure that a sufficient number of false enriched regions exist to obtain a good estimate of the FDR for binding sites, we set the LLR threshold to achieve a relatively high empirical FDR for enriched regions. We set the LLR threshold to 18.75 for the DosR dataset and 38.5 for the GABP dataset, achieving empirical FDRs for enriched regions of 0.389 and 0.40, respectively.


A blind deconvolution approach to high-resolution mapping of transcription factor binding sites from ChIP-seq data.

Lun DS, Sherrid A, Weiner B, Sherman DR, Galagan JE - Genome Biol. (2009)

Empirical FDR of CSDeconv. (a, b) The empirical FDR for enriched regions as a function of LLR threshold is shown for the (a) DosR and (b) GABP datasets. (c, d) With the LLR threshold fixed, the empirical FDR for binding sites as a function of the regularization factor α is shown for the (c) DosR dataset (LLR threshold at 18.75) and the (d) GABP dataset (LLR threshold at 38.5).
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC2812949&req=5

Figure 2: Empirical FDR of CSDeconv. (a, b) The empirical FDR for enriched regions as a function of LLR threshold is shown for the (a) DosR and (b) GABP datasets. (c, d) With the LLR threshold fixed, the empirical FDR for binding sites as a function of the regularization factor α is shown for the (c) DosR dataset (LLR threshold at 18.75) and the (d) GABP dataset (LLR threshold at 38.5).
Mentions: In Figures 2a and 2b, we show the empirical FDR for enriched regions as a function of the LLR threshold for the DosR and GABP datasets, respectively. We see that, owing to its lower coverage, larger LLR thresholds are required to achieve low FDRs in the GABP dataset. To ensure that a sufficient number of false enriched regions exist to obtain a good estimate of the FDR for binding sites, we set the LLR threshold to achieve a relatively high empirical FDR for enriched regions. We set the LLR threshold to 18.75 for the DosR dataset and 38.5 for the GABP dataset, achieving empirical FDRs for enriched regions of 0.389 and 0.40, respectively.

Bottom Line: We present CSDeconv, a computational method that determines locations of transcription factor binding from ChIP-seq data.CSDeconv differs from prior methods in that it uses a blind deconvolution approach that allows closely-spaced binding sites to be called accurately.We apply CSDeconv to novel ChIP-seq data for DosR binding in Mycobacterium tuberculosis and to existing data for GABP in humans and show that it can discriminate binding sites separated by as few as 40 bp.

View Article: PubMed Central - HTML - PubMed

Affiliation: Phenomics and Bioinformatics Research Centre, School of Mathematics and Statistics, and Australian Centre for Plant Functional Genomics, University of South Australia, Mawson Lakes Boulevard, Mawson Lakes, SA 5095, Australia. desmond.lun@unisa.edu.au

ABSTRACT
We present CSDeconv, a computational method that determines locations of transcription factor binding from ChIP-seq data. CSDeconv differs from prior methods in that it uses a blind deconvolution approach that allows closely-spaced binding sites to be called accurately. We apply CSDeconv to novel ChIP-seq data for DosR binding in Mycobacterium tuberculosis and to existing data for GABP in humans and show that it can discriminate binding sites separated by as few as 40 bp.

Show MeSH
Related in: MedlinePlus