How can i get only the upregulated and downregulated genes from chip-seq peak files

Question

[atsalaki]

I used R and I wrote a script that analyzes peak files from ChIP-seq experiments from ENCODE. I am new to this field and I want to find only the upregulated and downregulated genes so as to put them as input in a vizualization tool that will show relations between genes like these Gene A inhibits Gene B(A:->B) or GENE A triggers/regulates GENE B(A->B).

I need only the upregulated and downregulated genes .

These are some lines from the macs.anno file i created with my R script using ChIPpeakAnno package:

[atsalaki]

Thank you all for your time and the answers! I really appreciate all of you!

Answer 1

[ablanchetcohen]

Are you confusing RNA-Seq and ChIP-Seq data? You can get lists of upregulated and downregulated genes from RNA-Seq data, not ChIP-Seq data.

ChiP-Seq will give you the location of peaks, corresponding to the binding sites of transcription factors.

RNA-Seq will give you the expression levels of genes, allowing you to determine which genes are unregulated and down regulated between conditions.

[atsalaki]

What?? You mean i cant figure out which gene regulates the other and in what manner fron chip-seq experiments(e.g. HeLa-S3 cell type with STAT1(TF) chip-seq experiment)??

[Chris Fields]

Not directly, no. ChIP-Seq tells you where a particular TF may bind (or a histone modification may preferentially bind) and it may indicate, with variable experimental conditions, differential binding at a specific location, be it peaks called using various tools or fixed locations like TSS. And you could look for binding motifs for your TF in enriched regions.

But it does not indicate directly whether that gene's expression has changed, only that the TF's binding at that location appears to change. Pretty much every experiment I have seen published where this was performed needed gene expression data for validation; you would need to perform a followup RNA-Seq or at the very least a well-constructed targeted gene expression analysis, the latter with proper controls including genes similarly derived from the ChIP-Seq that show no effect.

EDIT: Note I didn't even mention how one ties peak calls to a single gene/TSS, which can be a bit subjective (is 500bp close enough? 5000bp? 50000bp? up and/or downstream?).

Answer 2

[ablanchetcohen]

With ChIP-Seq, you only get the binding sites of transcription factors. The link with gene regulation is indirect. You obtain only the list of genes that could be regulated by the transcription factor under study.

The analysis you seem to be attempting to do is more typical of RNA-Seq, where you obtain the expression level of all genes. From RNA-Seq data, you can get a list of genes that are unregulated or down regulated between the conditions being compared. You can then feed these lists to pathway analysis software programs, such as Ingenuity Pathway Analysis or ReactomeFIViz, allowing you to hypothesize about the regulation between the genes.

EDIT: Sorry this was supposed to be a reply to the comment from atsalaki. I wish Biosters had a delete option, at least for the 5 seconds after posting.

[atsalaki]

ChiPArray 2 is doing what i want to do i think...http://www.ncbi.nlm.nih.gov/pubmed/25916854 or am I wrong?

[ablanchetcohen]

ChIPArray2 integrates several type of omics data to generate gene regulatory networks, including transcriptomics or RNA-Seq data.. You can't build a gene regulatory network with ChIP-Seq data alone, with ChIPArray2 or any other software that I know of. At least, it's not the usual way of doing it. With ChIP-Seq data, you don't have any direct information about the expression levels of genes. ChIP-Seq data only determines the transcription factor binding sites. The transcription factor binding sites are of course related to the expression levels of the genes, but the relationship is not direct. You can't normally get lists of unregulated and downregulated genes from a list of transcription factor binding sites alone. If you also have the RNA-Seq data, you can do the correlation with the transcription factor binding sites.

Answer 3

[SP]

First, you would like to take peaks which are close to TSS, in your case I think insideFeature = inside AND lets say shortestDistance <= 2000bp you can do the same with upstream peaks with shortestDistance < 5000bp. Assuming shortest distance is shortest distance from nearest TSS. Genes annotated with these peaks can be easily used for deregulation analysis. Unfortunately, there is no golden rule. I hope this helps.

[atsalaki]

Yes the shortest distance is shortest distance from nearest TSS. With SD<=2000 i get the upregulated and with SD<5000 i get the downregulated . I dont understand why these gives me the up/down regulated genes!! Can you please explain if you dont bother please!

[SP]

I am late for the reply, as you already know now ChIP-seq does not give you up and down regulated genes, it only provides binding sites of your factor of interest (transcription factor or co-factor) basis on which you can look for gene's which are directly bound by your factor and potentially be directly regulated. Now if you have RNAseq or microarray data, you can search for regulated genes which are bound by your factor. To validate if your factor is responsible for these genes expression use some conditions to perturb your factor (e.g. knock-down, over-expression and knock-out of your factor ) and find if those genes are effected.

[atsalaki]

Is there any paper or link i could learn more about how to pertub the TF factor(e.g. knock-down, over-expression and knock-out of my factor ) and find if my genes are effected. Thanks all of you for the answers they really help me, iam so lost in this.

[SP]

This paper explains and compares multiple the differential chip methods Paper