Bed to BAM information transfer

Question

[rbronste]

Hi,

Wondering about the validity of converting bed to BAM, for instance in terms of TF foot printing...if lets say Ive done some analysis and gotten a consensus peak set in BED format, well after Macs2 peak calling, and I want to convert that to a BAM file for foot printing would this be a valid thing to do? How does that conversion represent or misrepresent read depth in the newly made BAM file vs one initially from Bowtie2 for instance. Thank you.

Rob.

Answer 1

[Petr Ponomarenko]

Could you please explain why you want to convert BED from MACS2 peak calling to BAM alignment file? How you envision this process from information standpoint?

BED file from MACS2 has information about peak start, peak end, summit position and level of enrichment. So for a big region of DNA, say, several kilobases long it will give you a few numbers.

BAM file has alignment information and consists of reads' data including quality, how they align to the genome and so on. So it is lots of data even for a few hundred nucleotides long genomic region.

[rbronste]

So the gist of the situation is this. I got some Macs2 peaks for a number of replicates (e.g. 4 individual BAM files) for an experiment and to get some differential peaks between conditions using DiffBind which basically creates a consensus Macs peak set and then finds your differential peaks. I want to do some TF foot printing on my data using Centipede and then see where I have decent binding within my differential peak sets. So the question is what module of information to use for the fooptprinting which requires a decent read depth, which while no individual BAM file offers, merging the BAM files does. The other option I was thinking of is as I asked, get the consensus peak set and make a BAM file out of that to footprint. Trying to figure out the most valid course. Thanks!

[Petr Ponomarenko]

Just to clarify, you assume there are two options for TF footprinting: 1) merge BAM files and run analysis, 2) run peak calling and use its result somehow in footprinting, ie to guide the way bam files are merged before footpring.

What experiment protocol was used? ATAC-seq? Is the standard bioinformatics protocol for your type of experiment working, or individual BAM files have way too little coverage (and this is what you meant in the reply)?

[rbronste]

Yes this is for ATAC-seq and basically I initially did not want to merge BAM files for peak calling and downstream analysis such as DiffBind and EdgeR etc however yes the individual BAM files on their own dont have enough depth for foot printing so looking for a way to reconcile the two. Basically footprint from merged BAM files and intersect those footprints with my consensus and differential EdgeR peak sets.

[Petr Ponomarenko]

Since you are using ATAC-seq for TF footprinting, then merging of BAM files should be straightforward solution, since you want to find little areas with absolutely no or very little binding, righ? Then you can use consensus peak calling from each BAM to find promoter regions and mark which putative TFBS from the first step are in promoters.

[Petr Ponomarenko]

But it would be very helpfull to hear from a person who had the same issue and solved it.

Bioinformatics... We have to deal with low-quality data so frequently at our work, that we can rename our specialty to "big crapy bio data scientist"

[rbronste]

In this particular case its not so much crappy data just sparse due to the small cell population Im working with, I guess not uncommon with ATAC in general.

[Petr Ponomarenko]

It was not meant to offend you in any way

[simone.thair]

Hi there, I am trying to do a similar thing- did you get this figured out and if so, how did you do it?