Hello everyone!
We are trying to use MACS for detecting enriched regions in a histone modification ChIPSeq experiment. I know that MACS is rather designed for transcription factors and sharp peaks, we will try different software tools, but currently we'd like to produce reliable results with MACS.
Quite a few users adviced to use MACS for histone marks with switching off the local background modelling, with the --nolambda True setting, and use a global background instead. This recent paper in Current Protocols in Bioinformatics also describes the same:
http://onlinelibrary.wiley.com/doi/1...0214-prot-0002
We tried MACS with different settings on a large number of samples, and when we switched off the local lambda, the result was a tremendous number of peaks, sometimes even ~20 times more than in default mode!!!
Examining the results we have seen that the histone marks are detected as many small peaks. Like the one you can see in the attachment file.
So I got the feeling that in this mode MACS generates a lot of false peaks (many peaks are reported where there should be only one). But people seem to be using MACS in this way nevertheless.
My questions are: have you ever encountered this issue? How did you dealt with it? Or splitting histone marks into small peaks is not a problem? Why people keep switching off local lambda when using MACS for histone modifications?
Thanks for any help or comment!
We are trying to use MACS for detecting enriched regions in a histone modification ChIPSeq experiment. I know that MACS is rather designed for transcription factors and sharp peaks, we will try different software tools, but currently we'd like to produce reliable results with MACS.
Quite a few users adviced to use MACS for histone marks with switching off the local background modelling, with the --nolambda True setting, and use a global background instead. This recent paper in Current Protocols in Bioinformatics also describes the same:
http://onlinelibrary.wiley.com/doi/1...0214-prot-0002
We tried MACS with different settings on a large number of samples, and when we switched off the local lambda, the result was a tremendous number of peaks, sometimes even ~20 times more than in default mode!!!
Examining the results we have seen that the histone marks are detected as many small peaks. Like the one you can see in the attachment file.
So I got the feeling that in this mode MACS generates a lot of false peaks (many peaks are reported where there should be only one). But people seem to be using MACS in this way nevertheless.
My questions are: have you ever encountered this issue? How did you dealt with it? Or splitting histone marks into small peaks is not a problem? Why people keep switching off local lambda when using MACS for histone modifications?
Thanks for any help or comment!