Overclustered sequencing (low PF, low Q30) V2 2x250

[hcarrington]

Over the weekend I set up a V2 2x250 16S sequencing run on our Illumina MiSeq (I work in a Core lab). I came back on Monday and the results were terrible. Overall Q30 was 58.91, it overclustered at 1031 (we aim for 700-800ish) and Cluster PF was 49.72.
Anyway, I repeated the sequencing run and actually gel extracted the library pools just to be safe, quantification was good and I used a subnanomolar protocol to load the cartridge. I loaded 4pm of library with a 10% spike of PhiX.
So I checked the results this morning and they are actually worse! Cluster density is 931 with cluster PF at 20% and Q30 for read 1 is 85%
Typically we have excellent results: cluster PF should be between 80-90%, Q 30 used to be really good, though lately we've had some trouble with our read 2 results dipping below 80% (has anyone else seen this in the past 6 months?)
But this situation is really terrible and I wondered if anyone has any advice/has experienced it before?

[fanli]

Are these amplicon libraries? It sounds like you have low diversity in your initial few bases, which is giving issues in clusters PF. Perhaps try more PhiX? Although we routinely do amplicon sequencing with no PhiX and get >80% PF.

Quality dropoff on R2 in the longer chemistries has been reported (confirmed w/ Illumina tech support).

[microgirl123]

Call tech support. Last time we had results like that, the optics needed to be adjusted.

[hcarrington]

They are amplicon libraries; we've always done 10% PhiX, but maybe that should be increased. I suspected something may have happened with the instrument, I guess I better ask for a PM from Illumina. So the quality dropoff in R2, is it a reagents issue? I noticed that the Lot number is the same for our last 5 V2 500 runs and they have had particularly low Q 30's. Is there a reason why this is happening, because this wasn't an issue last year.
Thanks everyone for the quick responses!

[GenoMax]

This thread is being referred to obliquely in posts above. That is for V3 reagents though. Don't think V2 reagents are affected.

[pmiguel]

Adding more phiX won't help. If you are at zero complexity for some number of bases (and you haven't said that is the case) then you just have to get that cluster density down to get a good result.

We actually had a very similar result with a 500 cycle run recently. We had a collection of amplicons from a customer but they also had another genomic DNA library they wanted as 30% of the run.

We clustered at a final concentration of 10pM and got a cluster density of 1170 cfu. Note that this was a fairly complex mixture of amplicons and a genomic DNA library. PF% was 31% and the run was only 54% >=Q30. So I called tech support and they replaced the reagents. This cluster density was within spec for a v2 kit, so we should have been okay.

Nevertheless we backed off on the concentration to 8pM, diluting from the same 20pM denatured stock that we'd used for the previous run. We got the expected decrease of 20% in number of clusters, to 979. PF% was somewhat better, 55%, but the quality was much better -- 70% >=Q30.

The second run was actually performed on a different MiSeq. So, the weirdness was that we were getting higher than expected cluster densities + I guess either the reagent cassettes and/or the MiSeq software are turning in lack-luster performances of late.

--
Phillip