Skip to content

How to compare transcription factor binding between two ChIP-seq samples

August 7, 2011

This is a question people seem to be having some difficult with, as I’ve seen it asked a few times on SeqAnswers. You have results from two ChIP-seq experiments.  For example, you want to know if NF-kB is recruited to different promoters in response to Sendai virus and rhinovirus.  So you in infect cells with Sendai virus and rhinovirus and do ChIP-seq with them using an anti-NF-kB (p65) antibody. You get back the results, run them in your favorite peak-finder and get a list of NF-kB binding sites under both conditions.  You then compare the lists and you find a lot of differences.  However there is a big problem here.  Small differences in the calculated p-value by the peak-finding program can lead to peaks being called or not called.  It doesn’t matter if you enter relaxed or stringent criteria into the peak-finding program.  You will always have this arbitrary line where some peaks will fall on one side and others on the other side due to small differences.  This transformation of a quantitative variable (p-value, fold enrichment, etc.) into a nominal variable (peak called or not called) will cause you problems when you compare the lists.

I suppose you could (and I’m sure someone will) write a program that takes into account differences in p-value, fold enrichment, etc. when calling peaks, but if you are like me, this would require too much programing knowledge.  An easy way around this is to use one sample, for example the NF-kB ChIP after Sendai virus infection ChIP as the control and the other say NF-kB after rhinovirus infection ChIP as the treatment sample.  This will give you the places where the tag density in the NF-kB after rhinovirus infection ChIP is higher than tag density of the NF-kB after Sendai virus infection ChIP.   Run the analysis the other way around to find opposite.

However, there is one big problem.  The input samples are not the same from the two conditions, so you may have identified regions that have a higher tag density due to chromatin structure and not NF-kB binding.  To get around this you should identify the actual NF-kB binding sites from your two samples as you normally would using your favorite peak-finding program.  So now you have four files (hopefully in BED format):  1) NF-kB binding sites in Sendai virus treated cells, 2) NF-kB binding sites in rhinovirus treated cells, 3) Areas of enriched tag density in Sendai Virus treated cells compared to rhinovirus treated cells, and 4) Areas of enriched tag density in rhinovirus treated cells compared to Sendai virus treated cells.  Take file number 1 and intersect it with file number 3 and you will have a list of places where NF-kB binds in Sendai virus treated cells and not rhinovirus treated cells (you can use Bedtools or Galaxy to do this).  Do the same for files number 2 and 4 and you will have a list of where NF-kB binds in rhinovirus treated cells and not in Sendai virus treated cells.

Update:
There is a new Bioconductor package for assessing differential binding in ChIP-seq data called DiffBind. I’m sure it is much more refined than my hacked together methodology above.

I haven’t tried it but it looks interesting. Give it a try, but if you are not familiar with R, then you need to read ‘R in a Nutshell’ first. But to understand ‘R in a Nutshell’ you really need have some understanding of scripting in UNIX, Perl or something.

Advertisements

From → Posts

3 Comments
  1. Recommendations for Differential ChIP-seq analysis can be found in the paper:

    Bailey T, Krajewski P, Ladunga I, Lefebvre C, Li Q, et al. (2013) Practical Guidelines for the Comprehensive Analysis of ChIP-seq Data. PLoS Comput Biol 9(11): e1003326. doi:10.1371/journal.pcbi.1003326

    http://www.ploscompbiol.org/article/info%3Adoi%2F10.1371%2Fjournal.pcbi.1003326

  2. ethanomics permalink

    This post was an quick hack from the old days when there wasn’t much around. I’ll have to take a look at the paper. For anyone with replicates, I’d recommend looking at the software package DiffReps.

Trackbacks & Pingbacks

  1. How to compare transcription factor binding between two ChIP-seq samples | BJC俱乐部

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: