Open MalindaH opened 4 years ago
The sequencing data that i'm using are double-strand break sequences. Does that have an impact in this output?
@MalindaH q-value is more conservative than p-value. You can see that in your first result, most of the peaks have -log10 p-value of 1.9 or 3 which means they barely passed the cutoff of 0.05. I suspect that those weak peaks are noises in the system. You can ask MACS to generate signal files (-B) in bedGraph format, load them to genome browser such as IGV or UCSC genome browser to see if those peaks look real. Another option is to run macs callpeak --cutoff-analysis
then you will see how the total number of peaks change with cutoff value, which will help you decide a good cutoff. You will have an insight that what extent you lower the cutoff, the peak caller starts to capture noises (a sudden increase of peak number).
Dear Dr. Liu,
I am trying to use MACS2 on human double-strand breaks files.
When I set the cutoff pvalue to 0.05 (in test14), I could get a lot of peaks, but the -log10(qvalue) column in the output files are almost all 0. Here is part of test14_peak.xls:
When I set the cutoff qvalue to 0.05 (in test15), I was only able to get one peak. Here is test15_peak.xls:
It seems that it's not calculating the qvalues correctly because the qvalue cutoff is not really working. Could you take a look and point out what have I done wrong?
Thank you!
System: