Hello,

First thing first: thanks a lot for providing the community with such a useful ChIP-seq tool as DiffBind :)

It seems I am missing something in how the score column values are computed in the 1st step of reading in the datasets with the dba() function.

My design table is in a csv file, using peaksets derived from MACS2 xls output file (broad peak).

I ran the following command to read-in the design table and MACS peaksets:

mydba <- dba(sampleSheet = filename, peakCaller = "macs", peakFormat = "macs", scoreCol = 7)

this gives:

head(mydba$peaks[[1]])  # MACS2 input peaks coordinates
  chr  start    end X.log10.pvalue.
1   1 778427 778623      0.07637541

The same peak info line in the original MACS2 xls file looks like this:

chr    start    end    length    abs_summit    pileup    -log10(pvalue)    fold_enrichment    -log10(qvalue)    name
1    778427    778623    197    778528    19.00    16.93802    8.87306    14.41193    myfile

Basically, the "score", i.e -log10(pvalue) from MACS2 is 16.93802. 

While dba() give a 0.0763 as a score.

I thought that normalization/processing was performed later on with dba.count(), that the dba() step was only to load the data?

My question therefore is: how is the -log10(pvalue) computed by dba()?

Regards,

Hatice

University of Fribourg, CH

DiffBind does a simple re-scaling of peak scores, mostly so that scores from different peak callers can be compared on the same scale. The scores for a peakset are all divided by the maximum score for that peakset, which places them on a 0..1 scale.

Cheers-

Rory

Okey, thanks for the info dear Rory.

Best,

H,