Variable Insert sizes in targeted sequencing data

[meher]

Hi,

We have targeted DNA-seq data for 30 samples. As part of the analysis we have calculated the insert size distributions in two scenarios:

1. Using all the aligned reads

2. Using all the aligned reads within the targeted region.

In these two scenarios, the insertsize distribution is highly variable as shown in the figures attached.

Could anyone comment on this variable insert size, why is it highly variable and
which one is the ideal insert size to calculate structural variants?

Any suggestions are valuable. Thanks.

[bastianwur]

If it was from an assembly, then I'd guess a collapsed repeat.
Might as well be an artifact from the shearing, that this region just has properties which lead to smaller fragments (but that's a pure guess).

[Brian Bushnell]

Another possibility is that the region is short, or full of short contigs. If there are no contigs longer than X, the reads with insert sizes substantially longer than X won't map paired, only the short ones will.