Visualization of ChIP-seq data (mapped reads)

Question

[Wet&DryImmunology]

Hi I was trying to play with some published ChIP-seq data. I noticed that the aligned reads presented has extremely sharp peaks (attached figure A), so I cooked the original data to see whether I could get similar results.

bowtie2 -p 15 -x ~/Bowtie2Index/genome -U SRR561494.fastq.gz -S TregFoxp3ChIP_by_Rudensky.sam  >& out.txt &
samtools view -bS TregFoxp3ChIP_by_Rudensky.sam  > TregFoxp3ChIP_by_Rudensky.bam  &
samtools sort TregFoxp3ChIP_by_Rudensky.bam -o TregFoxp3ChIP_by_Rudensky.sorted.bam &
samtools index  TregFoxp3ChIP_by_Rudensky.sorted.bam TregFoxp3ChIP_by_Rudensky.sorted.bai &

convert the bam file into bw file using bamTobw.sh enter link description here and load the bw file on to the IGV browser.

but somehow I could not get sharp peaks like the published figure as shown here. Is the author use some specific method to visualize the mapped reads, or something wrong with my cooking of the data? if any idea, please let me know. Tsk!

Image

[jotan]

Could you link to the original study?

It looks like they might have applied some transformations, maybe corrected for input.

Edit:

Tracked down the paper. Here's the vague one-line descriptor that might explain the sharp peaks

"When plotting tracks for multiple ChIP-seq experiments for a single transcription factor, a multiplicative factor was used to normalize for enrichment, i.e., differing numbers of tags in peaks."

[Mike]

In this case it would be better to ask author about input parameters.

[Ming Tommy Tang]

How does your peak look like? For transcription factors, the peaks are usually very sharp. Note that Figure A the scale is pretty big (40kb).

[Wet&DryImmunology]

the upper track is directly converted from mapped bam (sam) file into bw by using bamTobw the lower track is converted from bdg file created by macs 14 peak calling.

[Mads Lerdrup]

I know that I use a different approach, but I tried to map with bowtie and import the FoxP3 ChIP-seq into EaSeq (default settings) to see how it looked like. And I think that the peak-containing structure is remade better than in the example that you posted above (see image). IMHO then the bins in your tracks look quite large considering that Ikzf2 is app 150 kbp - a rough guestimate suggests to me that each bin in the imaged tracks is app. 1.5 kbp. If that is the case then the narrow 200-300bp peaks would become 5-7 times lower compared to the background. I don't know if the binning comes from your bigwig or IGV - in the latter case then you should be able to zoom in and see the peaks near the TSS better.

However, I don't get as high peaks relative to background and noise as in the published figure, but that could be because I remove duplicated reads, due to different mapping settings, because EaSeq extends the reads to fragments, or if the authors did some background subtraction.

Answer 1

[Ming Tommy Tang]

As jotan1982 mentioned, I guess the Input signal is subtracted from the IP. If you make two bigwigs(one for IP, one for Input), you can get a subtracted one using https://github.com/Ensembl/WiggleTools The output is wig file. you can use UCSC utility wigToBigWig convert back to bigwig for visualization.