Question: Diffbind: Using bigwig tracks in place of bam files
0
7 months ago by
ram.n.krishnan0 wrote:

Hi

I am using diffbind to identify perform DBA in ChIPRX normalized ChIP-seq data (https://doi.org/10.1016/j.celrep.2014.10.018). Since the normalization information is contained entirely in the bigwig tracks, I would like to use these trackfiles in place of the bamfiles for DBA. Is it possible to do this in Diffbind?

best Ram

diffbind • 154 views
modified 7 months ago by Gord Brown590 • written 7 months ago by ram.n.krishnan0
Answer: Diffbind: Using bigwig tracks in place of bam files
0
7 months ago by
Gord Brown590
United Kingdom
Gord Brown590 wrote:

No, DiffBind (and DESeq2 and EdgeR) require (un-normalized) counts, not coverage files. The rationale (for DESeq2 specifically) is discussed here: http://bioconductor.org/packages/devel/bioc/vignettes/DESeq2/inst/doc/DESeq2.html#why-un-normalized-counts

Sorry...

• Gord

Gord, thank you for the quick response.

I understand that the RPKM type normalization used in RNA-seq is not acceptable for DESeq2 and edgeR, since the normalization factor depends on local genomic features. However, in the case of ChIPRX, we perform global normalization where the read count at every genomic locus is increased or decreased by a single factor. For example, if we obtain $n$ reads at loci1 and $m$ reads at loci 2, then ChIPRX normalization would yield $an$ and $am$ reads at loci1 and loci2, respectively. Since global normalization does alter the read statistics will it be possible to use the normalized data for DBA?

DiffBind won't accept coverage files, normalized or not, but if you can convert them to count tables, then you should be able to run DiffBind on it. Rory recently added a comment on using DiffBind with counts directly:

https://support.bioconductor.org/p/119168/#119355

Given that the counts will not respect the assumptions of the underlying DE analysis tools, it's not clear what the results will mean. Maybe somebody more sophisticated than me can chime in on that.

Alternatively, try just running it using the BAM files, and see what happens. Maybe DESeq2's and/or EdgeR's normalization will be enough to give you something useful.

Cheers,

• Gord