No difference in peaks between ChIP sample and control / Bad background noise on control for ChIP-Seq / Weak enrichment for ChIP-Seq

Question

[kai.chiam]

I am a complete newbie to bioinformatics. Currently doing ChIP-Seq on samples looking for enrichment at sites for increased/decreased expression of a retinoic acid receptor with an agonist/antagonist. I relied on online tutorials such as this one by abcam and Youtube videos.

The data I was given are paired-end reads. Because there are 15 samples and each went through multiple runs, I have 40 -fastq.qz file for each sample.

These are the parameters used on Galaxy:

1. Trimmomatic: Sliding window to cut bases above quality score of 30 and above AND drop reads that are below quality score of 30
2. Align using bowtie2 using default settings for paired-end reads
3. Filter SAM or BAM, output SAM or BAM files on FLAG MAPQ RG LN or by region: minimum MAPQ quality score of 30 and filter on bitwise flag > skip alignments with unmapped reads
4. Samtools sort using default settings
5. Samtools merge using default settings
6. MACS2 call peak: paired end BAM (BAMPE) & default settings

I initially used quality scores of 20 to trim bases and filter unmapped reads. The peak calling showed very little difference between the peak in the ChIP sample and the control. I then ran them through the plot fingerprint tool on Galaxy and they showed very weak enrichment. I was advised by a colleague to up the quality score to 30 to try to reduce the background noise on the control but that didn't work either.

There is a possibility that there were not enough wash steps done in doing the ChIP or perhaps there is something in the analysis steps that wasn't quite right. I'm not sure how to go about troubleshooting this.

Any advice would be welcome.

Images of the plotfingerprint tool results and peaks after peak calling on the USCS browser can be viewed here on Google Drive

[Friederike]

What types of enrichment were you expecting to see for the ChIP'ed factor, i.e. more along the lines of H3K27ac (= narrow and tall) or more along the lines of H3K9me3 (= very broad and more subtle). The reason I'm asking is that depending on the type of signal you're expecting, it would be easier to declare the error to be on the wetlab or drylab side. See the example fingerprint plots for the different types of signal here.

[kai.chiam]

Hello, Friedrike. The type of enrichment expected is broad and more subtle.

Was MACS2 call peak a suitable tool for this? I know there is a MACS2 bdgbroadcall tool on Galaxy but as I am quite new to this, I am not certain of the parameters to be used and have not seen other posts/papers using this tool.

Thank you.