MACS: ChIP and control with different tag size

[dekontrol]

Hi,
I am using MACS2 for ChIP-seq peak calling.

I have a treatment file with original reads of 51bp. Before alignment, the reads were dynamically trimmed on 3'end for removing low quality bases (with Trimmomatic software). At the end, aligned reads ranges from 42 to 51 bp.

Then, I have a control (input) file with original reads of 36 bp. Before alignment, the reads were dynamically trimmed on 3'end for removing low quality bases (with Trimmomatic software). At the end, aligned reads ranges from 30 to 36 bp.

If I let MACS to automatically determine the tags size, MACS comes out with a wrong number. Moreover, since I think MACS scan only the first n reads, MACS does not give the average of the tags length.

So, I have to specify the tag size with the option --tsize. However, it is impossible to set a tsize for the treatment and a different tsize for the input.

What do you think is the best solution for this situation?

Thanks very much!
Albert

[dpryan]

Why didn't you sequence the IP and input at the same time? I'm guessing that you also didn't construct the libraries at the same time and therefore have an uncontrolled batch effect completely confounding your results.

[dekontrol]

In the real world, sometimes you have to use old data with new data, or data that comes from different labs.. any idea?

[SylvainL]

Even if I fully agree with dpryan, it happened to me that researchers really wanted to use their old datasets with a new one. Except that in my case, both samples were done with the same chromatin then fragments sizes were similar. I just trimmed the longest reads to fit the shortest...
Hope both samples (treatment and control) were obtained with the same chromatin (I mean same sonicated chromatin, not only from same celltype...)

If not, maybe the best would be to use MACS without control for both samples and then you crossed the results yourself...

[dekontrol]

Yes SylvainL, I am exacly in that situation: my wets wanted to use their old data.

Is there any paper that I can give to my wets, where is clearly stated that treatment and input should comes from the same chromatin?

Thanks guys for your help!

[dpryan]

They need a paper for that? They would have the exact same problem with their wet lab experiments. It'd be like taking an image of a Western blot from a paper and comparing the raw band intensities to one of your own Westerns. In fact, just use that example, since they should understand it.

[dekontrol]

They told me that they think that they can make only one input from the same cell type, and use it as a reference input for all the future experiments from the same cell type..

[dpryan]

Funny, so they must think that they only ever need to run 1 Western as a baseline for comparisons then too...

[SylvainL]

So typical when wet-biologists do not understand a technique... I had to deal with some of these quite recently. Then, we finished the discussion by me telling them: if you want to perform only one Input, then prepare a huge batch of chromatin... They didn't like this comment when they realized they would have to prepare 1000 plates together

If they don't understand the example of dpryan (they may argue that it depends of the exposure of the blot, etc...), simply ask them to summarize all the Ct they obtained for their control genes in qPCR (all same cell type of course) let's say the last 4 years....