I have two set of data about ATAC-seq and DNase-seq and I like to perform peak calling using MACS2, but I found that there are some differences in shift size (plus or minus direction). I like to know which one is better for either ATAC-seq or DNase-seq (e.g. --nomodel --shift -100 --extsize 200 or --nomodel --shift 100 --extsize 200). is it right that because most accessible regions in ATAC-seq and DNase-seq located in 5' region so using --shift -100 is recommended?
1 answer
Just use --nomodel alone. If using no shifting model (from my understanding and experience), the use of shifting parameters does not make sense and does not change the output in comparison to using --nomodel alone.
For ATAC-seq (paired-end, deduplicated) I like:
macs2 callpeak -t in.bam -n name --keep-dup=all --nomodel -f BAMPE
Update: My current preferred parameters are:
--extsize 100 --shift -50 --nomodel -f BED --min-length 150
where the input is a BED file with the cutting sites, so the start and end position of every paired-end fragment as 1-bp coordinates to force macs use all available information.
This has nothing to do with paired-end data. --nomodel deactivates the shifting model. Therefore, any shifting parameters do not make sense. I do not know why people use these settings, but it does not matter for the output. Try them out. You'll see (at least in my experience) that the number and location of the peaks will be the same. The use of BAMPE makes sense because in paired-end sequencing, one knows the true fragment sizes. When using BAM, macs estimates them, so BAMPE reduces the level of uncertainty.
--nomodel --shift 100 --extsize 200 is intended to assess Tn5 cut sites and represents cut site enriched fragments for peak calling analysis. Some of these parameters have been discussed. --nomodel does not exclude shifting it just doesn't build a shifting model. MACS2 documentation states that --shift and --extsize can be used together with --nomodel for the purpose of DNase-seq. In practice I do find that -f BAMPE provides more peaks than cut site enriched methods and we have validated these peaks with ChIP-seq datasets of transcriptional factors.
Log in to answer this question.
-
call peak for all replicates in all conditions using macs2
written by QX •
-
low FRiP(Fraction of Reads in Peaks) score in ATAC-seq
written by yoonho •
-
Using MACS2 parameters
written by Nishat •
-
ATAC-seq peak calling?
written by star
-
MACS2 option for ATAC seq?
written by star
-
When using MACS2 with -f BAM, it is possible to have a shift other than 0?
written by Lucy •
-
ATAC-seq macs2 peak calling bampe mode extension shift
written by nanoide
-
Peak calling using macs2?
written by star
-
ATAC-seq peak calling with MACS
written by igor
-
Help understanding MACS2 --extsize and --shift
written by datascientist28
