Mnase cause high multimapper fraction?

[Mkrause]

Hello Community.

I am working on histone modifications in zebrafish.
I recently ran several samples for a histone methylation ChIP after MNase digestion.

Now I run into the problem of having 60-80% of Multimappers in my mappable reads.
Does anyone have the same problems or knows why that could happen?

Some more specifications:
I have "produced" this problem in 10 independent samples, plus in 4 Input samples (meaning that it is not the IP that enriches for Multimappers).

Digestion by MNase was performed to have 80% Mono-and Dinucleosomes, but no over digestion (fragments smaller than 150bp)

FastQC gives OK quality. There are some ambiguities with Kmer content, but other than that quality seems ok.

I sequenced around 50Mio reads, from which 70-80% were mappable. However, from these, 60-80% were Multimappers. When analysing the Multimappers, they are distributed throughout the genome, and all "classes" of multiplicity is present.

From that I reason that already the MNase digestion prefers repetitive sequences, but I have no idea why and how to prevent this.

Does anyone has Input on this?

[epigenereg]

Did you try to report only uniquely mappable reads?

[Mkrause]

Yes sure.
IT makes 10-20% of the whole genome. However, in next experiments I do not want to waste 80% of my data on multimappers, if avoidable.

As the problem appears also in the Input, I suspect the MNase digestion or clean-up to cause the problem. This is why I ask for help

[Chipper]

What is your definition of multimappers?

[Mkrause]

That one single read maps to more than one position in the genome with the same confidence.

Is there any other definition?

[colindaven]

I analysed some public MNase data recently for a large complex genome >2gb and had this issue too with 50bp reads. There appears to be a very high background in MNase data.

Longer read lengths might partially mitigate the issue. However, if MNase indicates you are primarily sequencing complex repeats this problem will always occur if read lengths are shorter than the compound repeats. Interestingly, those using the data are very happy with it, especially the patterns around gene boundaries appear as expected.

To practically deal with it I have produced two tracks, one with all reads and one filtered by samtools to only include Mapping Quality 20+ reads.

[Mkrause]

Thank you all for your replies!

Would you think that using paired-end seq might help?
First of all it is likely that one of the two reads is not multi-mapper, and thus would give more confidence for the second read.
Second, since one knows the possible distance between the reads, one can further increase mapping confidence.

IS that an option?