Question: ATACseq with STAR and macs2
gravatar for Ice P
16 months ago by
Ice P0
United States
Ice P0 wrote:

I am new to ATACseq data. Earlier, I have processed mainly RNAseq and few Chip seq data, thus i used same tools for atacseq as well. I used STAR (default parameters plus --outSAMstrandField intronMotif) to generate bam and wig files, then I used macs2 (-B -q 0.01) to call peak using those bam files.
Is this a correct way to process? Also, can you recommend a tool to get the differential peaks from this ATACseq data?

star atac-seq dmrs macs2 • 1.7k views
ADD COMMENTlink modified 14 months ago by simon.vanheeringen170 • written 16 months ago by Ice P0

We use Bowtie for mapping, MACS for peak calling, IDR filtering of peaks and DESeq2 for differential peaks. Still work in progress but our current pipeline is on github.

ADD REPLYlink written 14 months ago by trausch1.2k
gravatar for ATpoint
16 months ago by
ATpoint11k wrote:

You typically use the --nomodel option, as the shifting model of MACS does not really make sense for open chromation data. As you have probably paired-end data, also use the -f BAMPE option. It forces MACS to pileup the real fragment length instead of an estimate, which maked sense imho, due to the quiet different fragment sizes that the library prep creates, also if you have paired-end why not make full use of it. Check reproducibility of the peaks between replicate samples, then rerun MACS with the merged bam file and feed the count matrix into DESeq2. Reference e.g. Corces et al 2016 Nat Genetics. Btw, any reason to use a RNA-seq aligner for DNA assays?

ADD COMMENTlink modified 16 months ago • written 16 months ago by ATpoint11k

Thank you. I started testing with STAR because it is very fast. I read that STAR could be used for DNA aassys coz it can handle gaps. Plus, using option like --alignIntronMax 1 --alignMatesGapMax <maxinsertsize-2*readlength> splicing could be prohibited. Anywa, seems I need to do everything again with bwa alignment and macs2. Btw, the reads are not paired end. Any further comments and suggestions are appreciated.

ADD REPLYlink written 16 months ago by Ice P0

Well, if single-end, use the same stratagey but without the BAMPE option. MACS will then estimate the fragment length and pile up the fragments accordingly.

ADD REPLYlink written 16 months ago by ATpoint11k
gravatar for simon.vanheeringen
14 months ago by
simon.vanheeringen170 wrote:

Have you tried the pipeline from the Kundaje lab?

Pretty much worked out of the box for me. For differential peaks you could do the following:

  1. merge all peaks into one peak set
  2. count reads for all experiments in this merged peak set (using bedtools multicov for instance)
  3. use the count table in DESeq/EdgeR
ADD COMMENTlink written 14 months ago by simon.vanheeringen170
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1740 users visited in the last hour