chip-seq profile using bigwig data

Question

[qazwsx]

Hi Guys, I am new to the field.I have a question about processing the chip-seq data using the bigwig file. In general, I wonder what is the standard way to plot chip-seq profile for a specific TF or histone modifications. Here is how I did it: I downloaded the bigwig data and transfer into bedgraph, which contains the number of reads in each uniform intervals and plot it in a profile. Basically I have the following two questions:

1. How to plot the continuous chip-seq profile like the ones in some published paper? As I am having all the concrete number of reads for each interval, e.g. 10bp or 25bp. Do I need to coarse grain the signal, like combine into a larger interval, like 10kb? Or do any kind of smooth?

2. How can I deal with the control file in this manner? I also have the bigwig file for the control but do I need to somehow subtract control from the target data? If so, what types of treatment does it need, like direct subtraction, or Poisson process to determine the significance? Or if someone can know if the bigwig for the target is simply ready to use.

Thanks in advance for any help. If there would be any idea I would be really appreciate!

Answer 1

[Devon Ryan]

You'll want to have a look at the deepTools package (fair warning, I'm one of the developers).

1. With deepTools, you'll want computeMatrix followed by either plotProfile or plotHeatmap. You don't need to change the bin sizes or smooth anything (the profiles will have their own bin size that you can adjust...I don't normally recommend smoothing).
2. You can either subtract that from the ChIP or take the log2 ratio (both are possible with bigwigCompare). I normally prefer to take the ratio, but sometimes the difference proves easier to interpret. Note that there's no automatic scaling done with this, so if these really do contain raw signal (i.e., they're not 1x normalized) then you'll probably have to figure out a reasonable scaling factor (in theory one can use plotFingerprint for this (it's not documented, but you can use bigWig files as input), but I expect manually fiddling with scaling until the input matches the ChIP's background will be easier for 99.999% of people). Regarding knowing what kind of information the bigWig file holds (i.e., whether it's already been normalized and is, thus, ready to use), that's hopefully listed in the metadata from wherever you downloaded it.