Base call pairs drifting apart over the course of a MiSeq run

[spamcorndog]

Hello everyone,

We've been having problems with our base calls, especially A and T, diverging over the course of a run (see attached screen shot). The runs are whole genome sequencing runs of bacterial genomes using Nextera XT, v3, 2 x 300 PE.

Obviously this results in all sorts of error generation, and we're having difficulty figuring out what's happening. We've spoken with tech support and so far haven't heard anything, and have spoken with several other people at Illumina as well who also haven't been able to help. Any thoughts on what could be causing this and what to do about it?

thanks!

[nucacidhunter]

The issue seems to be presence of library fragments with short inserts resulting in sequencing through adapters. Library need to be cleaned with beads for second time (0.5x) to remove smaller fragments.

[Brian Bushnell]

I agree with the diagnosis. You can confirm this by plotting an insert-size histogram based on read overlap, for example with BBMerge:

bbmerge.sh in1=r1.fq in2=r2.fq ihist=histogram.txt reads=1m

I suggest this because you it's hard to get an accurate insert-size histogram using mapping for reads that are much longer than their insert size.