Files to compare ChIP vs Input on genome browser: after alignment or peak calling?

Question

[salamandra]

When visualizing ChIP-seq data on genome browser should we use files obtained upon read alignment (bam files) or instead use the files treat_qvalue.bdg and control_lambda.bdg (from macs2) provided upon peak calling?

Answer 1

[Satyajeet Khare]

You can use treat_pileup.bdg.gz files and peaks.bed file. Bam files are not normalized to sequencing depth and are bigger in size.

[salamandra]

so treat_pileup.bdg is for chip, but then for input do i use peaks.bed? cause I though .bed didn't give 'read intensity'...

[Satyajeet Khare]

peak.bed file tells you which peaks are "real". There is only one peaks.bed file and it contains the co-ordinates of the statistically significant peaks.

[salamandra]

thank you, but I wanted chip and input to be different tracks on genomebrowser, which should I use for chip and which for input? do you know?

[Satyajeet Khare]

You can use treat_pileup.bdg.gz for ChIP and control_lambda.bdg.gz file for input.

[salamandra]

ok then if wanted to show only one track with the difference in intensity of ChIP minus Input I would use one of these two files: treat_pvalue.bdg or treat_qvalue.bdg files? or are those two files the ChIP intensity, but just for the significant peaks?

[Friederike]

for this, you could use deepTools bamCompare

[salamandra]

but still in that case, bamCompare doesn't select significant peaks, i'd still have to select those significant peaks based on macs2 peak.bed file, as sugested by Satyajeet?

[salamandra]

also, do we still have to remove duplicates when using bamCompare?

[Friederike]

yes, deepTools does not do any statistical assessment, that's what the MACS2 output is for.

[Satyajeet Khare]

You can show reads only from treat_pileup.bdg.gz file for regions where you observe significant peak. The peak is called at a region because no significant reads were observed in Input sample in that locus.

Answer 2

[Friederike]

deepTools' bamCoverage will allow you to generate bigwig files that are normalized for sequencing depth, i.e., they will be smaller in size than the bedgraph files and you can somewhat judge the differences in the peak magnitudes between different samples visually.

You could run bamCoverage for the ChIP and input BAM files separately and you will obtain bigwig files that you can easily load into a genome browser.

[salamandra]

But in that case would I have to first remove duplicates? I'm asking because I am not removing sequence duplicates after alignment to genome, as during peak calling MACS2 automatically removes them. But if use bigwigs got directly from bam files they will have duplicates right?

[Friederike]

if the duplicates are marked as such, e.g. after PICARD's MarkDuplicates, they can be ignored using one of the many options of bamCoverage