This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Calling differential peaks on large set of 3 conditions

Hi all,

What are the recommended methods to call differential peaks for enrichment/pulldown datasets (e.g. MEDIPS, H3K4me3) of different conditions with many replicates?

  • 20-30 biological replicates of condition1 pulldown+input pairs
  • 20-30 biological replicates of condition2 pulldown+input pairs
  • 20-30 biological replicates of condition3 pulldown+input pairs

So far I have seen the following:

This gives me the right output, e.g. regions enriched in condition1 vs condition2 and vice versa, as well as common.

The question is how do I combine my 20-30 biological replicates of condition1 and my 20-30 biological replicates of condition2 in macs2 bdg files to run the bdgdiff pairwise comparison mode?

https://omictools.com/polyphemus-tool https://omictools.com/differential-identification-using-mixtures-ensemble-tool https://omictools.com/mmdiff-tool https://omictools.com/genome-wide-generalized-additive-model-tool https://omictools.com/normr-obeys-regime-mixture-rules-tool https://omictools.com/chromstar-tool

  • Other tools I discarded for this purpose:

    • Segway, ChromHMM, ChromImpute: don't seem to have a differential mode.
    • Epicode: differential mode seems to give differential "codes" as output, instead of regions (peaks).
    • PARE: seems to call differential regions with no nucleosomes, rather than differential peaks.

Thanks

peak calling differential macs2
• 4,800 views • link
updated by Fluorine Fluorine • written by 14134125465346445 14134125465346445

1 answer

I would recommend DiffBind R package which in my hands works best for MEDIPS or ChIP-seq analysis with replicates and/or multiple conditions. MACS2 bdgdiff also works well, but then I would do an additional step afterward to see which are the common peaks between replicates, which you can do with IDR analysis. You can combine your samples by combining alignment SAM or BAM files with samtools merge.

THOR should, in theory, do the same as DiffBind but I don't have much experience with it.

• 0 views • link
written by Fluorine Fluorine

DiffBind is basically a wrapper around DESeq2 with parameters optimally tuned for ATAC-seq data, i.e. based on how you set up the contrast it will use the replicates to gauge the variability per condition.

csaw is a similar wrapper around edgeR, which also offers the same type of syntax for specifying arbitrarily complex comparisons.

• 0 views • link
written by Friederike Friederike

FYI DiffBind uses both DESeq and EdgeR, based on user's choice.

• 0 views • link
written by Fluorine Fluorine

Log in to answer this question.

More posts like this