Question: Is the MACS pileup output normalized
gravatar for ivoryec
4.5 years ago by
United States
ivoryec60 wrote:

I've read that when you give MACS an IP file and an input file (-t and -c), the larger library is scaled to the smaller library.

My question is: Do the bdg files (pileup and lamda) reflect this scaling?

I have samples from different conditions and I would like to show the graphs side by side but the library sizes were quite different. Each sample has an IP and an input file, so each sample was run through MACS independently.
My understanding is that each pileup can be considered to be normalized for library size in comparison to the corresponding lambda, but not between different IP samples.... is that correct?

pileup normalization macs • 3.5k views
ADD COMMENTlink written 4.5 years ago by ivoryec60

I posted this question to MACS github issues and received this response:


ADD REPLYlink written 4.5 years ago by ivoryec60
gravatar for Manvendra Singh
4.5 years ago by
Manvendra Singh2.0k
Berlin, Germany
Manvendra Singh2.0k wrote:

I guess

No, Its not normalized by library size but  its background corrected file, but you can normalize it afterwards, if you are doing statistics on them.(just divide the average tag count by number of reads mapped in million)

ADD COMMENTlink modified 4.5 years ago • written 4.5 years ago by Manvendra Singh2.0k

I didn't think the pileup graph took into account the background (input) sample. When I compare them, their ratio at the summits is almost exactly equal to the fold enrichment.  I thought I would have to integrate the two graphs (pileup and lambda) to get a background corrected file.

I was considering making a new graph of the ratio between the pileup and the lambda for each sample and using that to visualize comparisons between the samples.  Would that be wrong?

ADD REPLYlink written 4.5 years ago by ivoryec60

pileup is shifted signal, you can divide your window into smaller bins, then count the number of tags in each window.

I am not sure about ratio, that it would tell you anything, for this you can make a threshold and just look into peaks

ADD REPLYlink written 4.5 years ago by Manvendra Singh2.0k
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 832 users visited in the last hour