The distribution of peak in ChIP-seq on function element

[gigigou]

I have got some peaks from a ChIP-seq data.
Now I'm trying to locate these peaks on the function elements.
But I come up with a question while doing this.
That is a peak may be extreme long that it covers many function elements.
for example, let ~ represents peak, === represents exon, ------ represents intron

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
=======--------------================-----------=========

In this case, which element should I chose?
Thank you!

[mudshark]

if it covers introns and exons it basically covers genes, your feature definition might be too fine grained.

[dariober]

Originally Posted by gigigou

In this case, which element should I chose?

I don't think there's a clear cut answer to this question. It probably depends on the particular mark you are looking at (histones, transcription factors, methylations, etc...) and on the particular gene or genomic region as well as the technical resolution you can achieve (your peaks can be very wide because of the limitations of the technique and/or beacuse the mark is really that large.)

Searching for a one-to-one relation between peak and genomic feature will make downstream analyses and interpretations easier. But does it make sense biologically? I guess it is perfectly possible for a peak (histone, transcription factor...) to span and control several elements (genes, promoters etc...).

If you could give more detail about your experiment I think you could get better answers, especially from biologists.

Good luck!
Dario

[mudshark]

if there is a clear peak summit try to use this position for getting a one-to-one relationship

[gigigou]

Originally Posted by dariober

I don't think there's a clear cut answer to this question. It probably depends on the particular mark you are looking at (histones, transcription factors, methylations, etc...) and on the particular gene or genomic region as well as the technical resolution you can achieve (your peaks can be very wide because of the limitations of the technique and/or beacuse the mark is really that large.)

Searching for a one-to-one relation between peak and genomic feature will make downstream analyses and interpretations easier. But does it make sense biologically? I guess it is perfectly possible for a peak (histone, transcription factor...) to span and control several elements (genes, promoters etc...).

If you could give more detail about your experiment I think you could get better answers, especially from biologists.

Good luck!
Dario

Thank you very much.
I'm trying to build an analyze pipeline for ChIP-seq data. It doesn't focus on a specific experiment. So I have to figure out a solution which can get the one-to-one relationship.
Maybe mudshark's suggestion can be a instruction.

[gigigou]

Originally Posted by mudshark

if there is a clear peak summit try to use this position for getting a one-to-one relationship

Thank you!
I'll try your suggestion.
But I have a question, this will lose much infomation, doesn't it?
The summit should be highly credible as it represents the whole peak.

[DZhang]

Please be aware of the dangers using a fixed pipelines for different experiments. As it has been said by the previous posts, different biological questions may require different choice of features. I'd build several options in the pipeline to handle a few common scenarios. Most important of all, talk to whomever is going to use your pipelines first before making a decision.

[gigigou]

Originally Posted by DZhang

Please be aware of the dangers using a fixed pipelines for different experiments. As it has been said by the previous posts, different biological questions may require different choice of features. I'd build several options in the pipeline to handle a few common scenarios. Most important of all, talk to whomever is going to use your pipelines first before making a decision.

Your words are absolutely right. I'm going to set some options.
BTW, is there any tool finding genes according to peak coordinates?
I'm currently using PeakAnalyzer, but some of the genes found by it cann't be recognized by GoTermFinder.
Thank you!

[dariober]

Originally Posted by mudshark

if there is a clear peak summit try to use this position for getting a one-to-one relationship

In addition you (gigigou) could set a "priority cue" of genomic features to further restrict the annotation. This list could (and should) be defined by the end user of the pipeline. For example, a cue roughly suitable for a transcription factor could be:

TSS > 5'UTR > exon > 3'UTR > intron > intergenic

So if a peak overlaps equally well a transcription start site (TSS, possibly extended upstream by n bases) and an intron, choose the TSS as annotation.

Dario

[DZhang]

gigigou,

As to PeakAnalyzer vs GoTermFinder, I am not sure if the problem resides with your annotation, PeaAnalyzer or GOTermFinder. You need to examine the genes that cannot be recognized by GOTermFinder more carefully. As always, please make sure your gene annotation and GO terms match. (An old gene annotation may not work well with GO terms with the latest gene annotation.)

Regards,
Douglas
www.contigexpress.com

[gigigou]

Originally Posted by dariober

In addition you (gigigou) could set a "priority cue" of genomic features to further restrict the annotation. This list could (and should) be defined by the end user of the pipeline. For example, a cue roughly suitable for a transcription factor could be:

TSS > 5'UTR > exon > 3'UTR > intron > intergenic

So if a peak overlaps equally well a transcription start site (TSS, possibly extended upstream by n bases) and an intron, choose the TSS as annotation.

Dario

This is an excellent idea.
I think I need to figure out the "priority cue " first as I'm a green fingure in ChIP-seq.
Thank you.

[gigigou]

Originally Posted by DZhang

gigigou,

As to PeakAnalyzer vs GoTermFinder, I am not sure if the problem resides with your annotation, PeaAnalyzer or GOTermFinder. You need to examine the genes that cannot be recognized by GOTermFinder more carefully. As always, please make sure your gene annotation and GO terms match. (An old gene annotation may not work well with GO terms with the latest gene annotation.)

Regards,
Douglas
www.contigexpress.com

I have checked the unrecognized genes on GO website. They are indeed exist. And the gene annotation database used by GOTermFinder is updated as well. I'll look into this problem in other aspects.
Thank you very much!

[DZhang]

Please share your findings in the forum.

Regards,
Douglas
www.contigexpress.com