MiSeq Run Failed: No base calling

[CristianNW]

Hi everyone,
We are working with 18S Meta2 primers for metazoan communities in the soil, peat specifically. Our first run has failed completely after read 1. Global Q30 is 44%, but read 1 is >70%, while the rest are <51%, and read 4 is ca. 15%!

It is the first time we work with this particular marker, but we have gone through all quality controls during library preparation and have no reason to believe there are any particular issues with the library. The final product after indexing (PCR2) is 750 bp, which is substantially longer than our MiSeq can sequence, but again, we have no reason to think this would cause this issue. Q30 plummets after roughly 200 cycles, seemingly due to an incapacity to call bases. Which rises steadily up 5% at around 300 cycles, and after that goes 'wild'.

We considered whether overclustering could be the issue, partly thinking that the length of the marker may be the issue, but troubleshooting with Illumina technicians overclustering seems to not be an issue, with a density of ca. 600, and no issues on our PhiX. We have considered whether the reagents may have been the issue (Flowcell 20873144; PR2 20869085; Reagents 20877577). However, we are now waiting in a second run (216 cycles in) and although indices are better, % no call is rising steadily and is currently sitting at ~2.5, and Q30 oscillates around 75%. Unsurprisingly, the signal to noise ratio droped steadily too.


Both sets of reagents came from the same lot (may the whole batch be compromised?). We loaded both runs at 8 pM at 10% PhiX. We have also discussed whether the marker length could be an issue, but have been told that it should not trigger such a failure as both reads are independent. The indices used have been used before in successful runs too.

We had an earlier run fail in a similar way, with a mixture of markers (16S, COI1 Leese, Leray, 12S Kelly), but re-running with a slightly lower concentration seemed to fix the issue, hence why twe thought overclustering could have been the issue in our current run.

Has anyone had similar issues? Or has any indication as to what could be the source of the problem?

Many thanks,
Cristian.
Edit: The second run has marginally better figures but has failed in the same way. We have tried to find potential sources of high no call %s but other than under or over loading a run, which as per the density seems not to be an issue.

[CristianNW]

I thought I'd follow this up. We have run all possible diagnostics andn there are no issues with the machine or the library, beyond the length of the marker. Our working hypothesis, is that at this marker length, the synthesis during the clustering phase may not be able to fully copy the amplicon, which may lead to only partial sequences of the sequencing adapters. This could prevent bridge formation and clonal amplification, and thus lead to a small proportion of forward reads available for sequencing, and an overall weak signal, and as noise increases (as it does naturally), the % of call base drops. Crucially, in this scenario both index reads, and reverse reads would be expected to fail, since bridge formation is impeded, which they do.
Lower loading concentrations and higher PhiX % may partly alleviate this by reducing the ratio of DNA to synthesis reagents, increasing the synthesis efficiency, and signal strength throughout the run.
It's a hypothesis. We will re-run the library with 4-6 pM and 20-30% PhiX, and see if it improves the quality of the run.

Best,
Cristian.