This is a test version of Biostars. For the public version, visit https://www.biostars.org.
SEACR and MACS2

Hello everyone,

I have done cut&run for a transcription factor in a mutant cell line (mutant for mentioned transcription factor as a control). I did peak calling using MACS2, as expected few peaks were called in both the replicate. When I repeat this using SEACR, in one of the replicate zero peaks were called and in other replicate some 60,000 peaks were called. I am not able to solve this discrepancy. I mixed spike-in DNA during sample preparation.

Please help me. Thank you.

macs2 seacr
• 7,462 views • link
updated by rfran010 • rfran010 • written by qudrat.nii • qudrat.nii

1 answer

I would recommend looking at the genome signal tracks to see if the reps make sense compared to control or if there is bad signal in any of the reps. It seems like at least one of the reps may not have worked. You may also take the SEACR peaks and run motif analysis to see if you TF binding motif is enriched as expected.

• 0 views • link
written by rfran010 • rfran010

Seconding the recommendation of @rfran010 to look at the data in a browser, that is no better QC than this, because it uses the best tool there is = your brain+your eyes.

as expected few peaks were called in both the replicate

Expectation would actually be a good overlap between replicates for a good experiment. If little overlap is expected then it means that either the antibody itself is poor, or one of two experiments had a bad batch. I would actually stick to MACS2. These other tools expect pristante data quality with very little to no noise, so something that never works once an antibody is involved...

Can you share some screenshots from the browser? It seems one replicate simply did not work -- that is not uncommon in ChIP/CUT&RUN-like experiments, I see this all the time. After all, it is such a noisy sort of experiment with all sorts of sources for bias, you need to smile at your cells in a proper way, dance for the ChIP gods, wait for the right distance to the moon, maybe a sunny day... Ok seriously, ChIP experiments are variable, sometimes a good protocol simply fails with no obvious reason. If this was the case and one replicate was fine consider to simply do it again.

• 0 views • link
written by ATpoint ATpoint

enter image description hereHave a look at IGV screenshot; first two is IgG and the next two is for transcription factor

• 0 views • link
written by qudrat.nii • qudrat.nii

Have a look at IGV screenshot; first two are IgG and the next two are transcription factor mutant

• 0 views • link
written by qudrat.nii • qudrat.nii

It's too wide. Zoom to a locus, such as Gapdh so one can actually see individual peaks. Make sure it's scaled to same height across tracks.

• 0 views • link
written by ATpoint ATpoint

The signal-to-noise ratio is low. The quality of the data may not be good.

• 0 views • link
written by Ming Tommy Tang Ming Tommy Tang

I unfortunately don't see any peaks. It's only noise. If this is representative for the dataset then you might not get much out of this.

• 0 views • link
written by ATpoint ATpoint

Actually it is mutant for transcription factor when I ran SEACR for both of them the first file produce few or no peaks but the second produce some thousands of peaks. Note: macs2 produced few peaks in both the files

• 0 views • link
written by qudrat.nii • qudrat.nii

To clarify, did you do CUT&RUN for a TF in a mutant line that has said TF deleted or a specific mutation?

• 0 views • link
written by rfran010 • rfran010

yes. Actually we want to check a drug efficiency so for negative control we did this.

• 0 views • link
written by qudrat.nii • qudrat.nii

Okay, so you want to see no signal then?

Your results make sense then? In my experience SEACR seems to call something a peak by looking at regions of signal surrounded by regions of depleted (lower than background) signal. The peaks called by SEACR you see may be because one rep has less coverage of the genome. This maybe because it has lower sequencing depth, or something happened during library prep that decreased complexity. To that note, I echo ATpoint to rely more on MACS2/3.

Also, as a note, if you don't have evidence that this experiment works in a wild-type line, or equivalent positive control, your lack of signal could be because of a bad antibody or some other problem with your protocol. I am not really clear on your design or goals, but it is something to keep in mind if you are making decisions or conclusions based on this experiment.

• 0 views • link
written by rfran010 • rfran010

we have done experiment in wild type and it worked. I just want to solve this SEACR issues. Any leads or links to get this problem solved will be highly appreciated.

• 0 views • link
written by qudrat.nii • qudrat.nii

How do the fingerprint plots look for all samples?

Did you do SEACR with IgG as background? I don't know if it downsamples automatically, so that might be a consideration.

A little redundant, but a good starting point is to view the called peaks and see if that can help explain why they were called.

And, I'm curious, is there a reason you have to use the SEACR results and not MACS2?

• 0 views • link
written by rfran010 • rfran010

fingerprint plots look good. SEACR has been especially designed for Cut&Run and MACS2 misses some peaks in positive control. I did Cut&Run with spike-in DNA added for normalization purpose so did not use IgG.

• 0 views • link
written by qudrat.nii • qudrat.nii

Looks good, as in looks like a straight line? And they are all completely overlapped?

• 0 views • link
written by rfran010 • rfran010

Alright and your tracks do show some differences between the two reps.

SEACR doesn't have a lot of options, so you could try to run IGG as background this time. Hopefully that would get SEACR to recognize those peaks as false positive, but otherwise you would need to get eyes on the called peaks to determine why they are being called.

• 0 views • link
written by rfran010 • rfran010

Log in to answer this question.

More posts like this