mapdamage 2.0 doesn't work with paired ends

[Nihil347]

Hi,

I'm working with adna and use mapdamge to rescale base qualitys specific to the damage patterns of adna.

Everything works fine for single end reads but with paired end reads I get this errormassage:

Error in aggregate.data.frame(as.data.frame(x), ...) :
no rows to aggregate
Calls: plot.length ... aggregate -> aggregate.default -> aggregate.data.frame
Execution halted
Error: plotting with R failed

It looks like a problem with R, but that could not be the case because everything works fine for SE, as mentioned above.

The mapdamage homepage says:

"We assume the pairs are facing inwards when counting the position specific misincorporations and rescaling quality scores. An additional assumption in the rescaling process is the pairs are non-overlapping. Make sure the pairing information in the BAM/SAM file is correct as we rely on the paired end information provided by the file"

http://ginolhac.github.io/mapDamage/#a4

So i tryed

-samtools view -f 0x02 -b in.bam > out.aligned.bam
-samtools view -h -F 4 -b in.bam > out.bam and
-clipOverlap to edit the bamfiles before I use mapdamage.

nothing helps.

any idea?

[ginolhac]

Hey Nihil347,


thanks for your post. Actually, there are two different things and what you observed is a desired bahaviour that is now handled gracefully.

The R error comes from the plotting of the length distributions. The plot.length only works for single-end reads. It could work in theory on paired-end using the length of each pair that we think this does not represent the real insert size and could be misleading.
With ancient DNA, we do not recommend to use paired-end data that are not overlapping as large pieces of DNA are likely to be contaminants.
But if you still want to use non overlapping paired-end data then you can and since
commit b0cb37c (15 July 2013) the program continues and print a warning about the length distribution and this pdf is not produced. The disclaimer you cite on the website refers to the orientation of paired-end reads and is not link to the length.

For the rescaling, this is totally different and is based on the mis-incorportion frequencies.
This feature was added recently by Hákon and you can see commit 523b922, 22nd August 2013, mapdamage tagged version 2.01.
The absence of paired-end length distribution does not interfere at all with the Bayesian inference and the rescaling.

Hi,

I'm working with adna and use mapdamge to rescale base qualitys specific to the damage patterns of adna.

Everything works fine for single end reads but with paired end reads I get this errormassage:

Error in aggregate.data.frame(as.data.frame(x), ...) :
no rows to aggregate
Calls: plot.length ... aggregate -> aggregate.default -> aggregate.data.frame
Execution halted
Error: plotting with R failed

It looks like a problem with R, but that could not be the case because everything works fine for SE, as mentioned above.

The mapdamage homepage says:

"We assume the pairs are facing inwards when counting the position specific misincorporations and rescaling quality scores. An additional assumption in the rescaling process is the pairs are non-overlapping. Make sure the pairing information in the BAM/SAM file is correct as we rely on the paired end information provided by the file"

http://ginolhac.github.io/mapDamage/#a4

So i tryed

-samtools view -f 0x02 -b in.bam > out.aligned.bam
-samtools view -h -F 4 -b in.bam > out.bam and
-clipOverlap to edit the bamfiles before I use mapdamage.

nothing helps.

any idea?