3

The occupancy of SMARCD3 in the target genes listed below. I want to see average, normalized ChIP-seq signal at the promoter proximal region (1000bp upstream and downstream of the TSS).

I have 4 different experimental conditions (overlayed in one plot, per target gene using different color codes) where we want to visualize the changes in occupancy of GATA1 under all the conditions at individual gene promoters. The following are our target genes

  1. ABCB1
  2. ABCC1
  3. ABCC2
  4. SNAI2

So how GATA1 occupancy is changed across all these target genes across 4 different treatment condition.

representative figure

So far what i have done

  1. Aligned the samples
  2. Done peak calling
  3. Annotated the peak

I did steps 2 and 3 using homer.

Now after all this how do I find the occupancy of a TF with its target genes? Do I get the information from my peaks annotated output or from the peak files?

kcm
  • 1,804
  • 12
  • 27

1 Answers1

4

I would ignore peak calling for this and instead compute enrichment of ChIP/input for the genome (e.g., with deepTools or presumably homer) and then plot it for the genes of interest individually (e.g., using IGV or pyGenomeTracks) or as a group (e.g., with computeMatrix). If the peaks are obvious and you trust your peak calling then sure you can just use that instead, but given that that's prone to error it's usually better to simply use normalized signal.

Regarding actually looking at occupancy, I don't think that's useful. Occupancy isn't binary. If you're instead wanting to associate peaks with genes then using the nearest gene is common (it's not great, but it's not terrible either).

Devon Ryan
  • 19,602
  • 2
  • 29
  • 60
  • actually i looked into your deeptool as many publications used to similar analysis but not sure how it was done. " Occupancy isn't binary. " yes i now get it occupancy can;t be binary , " If you're instead wanting to associate peaks with genes then using the nearest gene is common" how do i perform this part in terms of tool can i do it in deeptools? – kcm Jun 19 '19 at 19:40
  • 1
    For finding the nearest gene, just use bedtools nearest from bedtools. – Devon Ryan Jun 19 '19 at 20:42
  • https://www.sciencedirect.com/science/article/pii/S109727651930228X in this paper this figure (G) Metaplots of canonical PRC1 (PCGF2, CBX7, and PHC1) cChIP-seq at classical Polycomb chromatin domains (n = 2,096) in Pcgf4−/−;Pcgf2fl/fl ESCs (UNT and OHT), in this paper did they get into this? did they use bedtools nearest as well to find the gene as they write "recruits canonical PRC1 to Polycomb target genes in order to compact chromatin " – kcm Jun 20 '19 at 06:18
  • They didn't do any gene assignments in that subpanel, they just input known polycomb domains. – Devon Ryan Jun 20 '19 at 06:39
  • okay can you tell me how did they get the profile of particular genes in other words where do get this information for doing Metaplots? is it from the peak annotation – kcm Jun 20 '19 at 09:02
  • Those are from igv or pygenometracks or similar. – Devon Ryan Jun 20 '19 at 09:33
  • okay may be the data i have used the read counts are really low so i dont get that profile – kcm Jun 20 '19 at 09:40
  • I hope you are talking about this one G) Metaplots of canonical PRC1 (PCGF2, CBX7, and PHC1) cChIP-seq at classical Polycomb chromatin domains (n = 2,096) in Pcgf4−/−;Pcgf2fl/fl ESCs (UNT and OHT). – kcm Jun 20 '19 at 09:42
  • 1
    No, they didn’t need or use genes for subpanel G. – Devon Ryan Jun 20 '19 at 09:46