Hi gt

In the image you included, You can see that the red pairs have an insert size larger than expected. This MIGHT be due to the presence of a deletion in your sample, since the reads would be at the expected distance in your sample, but further separated in the reference dues to the extra part of the genome your sample is missing. The fact that there are overlapping reads with the expected insert size suggest that the deletion would be heterozygous.

Variant callers (such as freebayes) can detect only small deletions (up to ~30 - 50 bp, depending on the tool) and for larger events you'll need to use a structural variant caller (Lumpy, for example, but many others exist).

I'd suggest first to tell IGV to show you the soft clipped reads (https://software.broadinstitute.org/software/igv/Preferences). If a real deletion is present, you'll see an accumulation of soft clipped reads showing you the exact breakpoints of your deletion (and the actual soft-clipped bases should map to the other side of the deletion)

UPDATE: An example to clarify the soft-clipped reads

In this image (from this paper)

there are two samples with deletions* in the same region (CDKN2A) . The top one has two nested deletions. These stripes of multicolor bases are the soft-clipped parts of the reads (not shown by default in IGV, I reaaaally recommend activating them). In the places where there's an accumulation of soft-clipped reads, we can see many reads that match perfecty up to a point and then do not match at all (soft-clipped part). This marks the exact position of the breakpoint, and the soft clipped sequence should be the same for all reads and match the sequence at the other side of the opposite breakpoint.

Is it a bit more clear?

Bernat

(*) Actually, in this case these are really translocations, but the image you'll see for deletion will be basically the same.