Paired reads overlap "too much"

[Retro]

In Illumina paired-end libraries I repeatedly see pairs, where the forward and reverse reads overlap over the entire length plus the forward read even continues (e.g. 30 nucleotides) after the beginning of the reverse read (an vice versa the reverse read sticks over the beginning of the forward read). I am not sure how this can arise, is this an artefact? It is not a rare finding and we can see it even in public data from NCBI SRA archive.

[nucacidhunter]

It depends on insert size and number of sequencing cycles. Whenever sequencing cycle number exceeds insert size, R1 and R2 will overlap. If inserts are shorter than one of read's length, then adapters also will be sequenced.

[Retro]

Thank you. That is what I was thinking, but: 1. these sequences are not adapters and 2. adapters should be trimmed from the data.

[Chipper]

That is not possible. Probably your reads are untrimmed and the viewer shows the whole read, not only the aligned part. What comes after the overlap has to be adaptors.

[Retro]

Could it be that the start of the R1 and R2 sequences are not determined correctly (trimming may remove adapter plus part of non-adapter sequence by mistake)? Then the sequences might overlap with 3'overhangs?

[dpryan]

Dove-tailing like that usually only happens due to trimming. Sometimes reads start with a stretch of crappy calls and if the read size is large and the insert size small...

[pmiguel]

By the way, it is easy to construct libraries with larger inserts. Just use different shearing parameters and/or more stringent ampure cuts.

[bastianwur]

No chance that you have some repeats there?