[nucacidhunter]

In addition to read quality there are problems with consistency as well. Re-running the same library with the same input gives different cluster numbers. The problem definitely is worse with 16S amplicons. High diversity libraries read quality are better but not as good as they used to be or up to the specification.

[toneta2013]

It sounds to me like a diversity issue, since the lowest diversity will be near the beginning. Are you using staggered primers, or spiking in anything with more diversity? There's also cluster density; even if you follow the same protocol, a higher cluster density will suffer more from a lack of diversity.

The run previously shown was run with a 4 pM 16S-library, 10 % PhiX spiked in. This resulted in a cluster density of 869. Are there anyone out there running the Illumina metagenomics protocol? Could you please provide a snapshot of a Fastqc-report from a successful run?
We are in dialog with Tech Support that are constantly sending us replacement kits and try to convince us that there do excist kits that are working... But it is frustrating to repeat runs that we assume will give crap results. And as long as Illumina do not officially acknowledge their trouble, I guess we will not be told when everything are back on track again...

[james hadfield]

At Illumina's UK UGM the team discussed the low quality at the end of long-read MiSeq kits. They acknowleged there was an issue and that they were actively working on it. However they have no plans to issue a quality notification until they understand the root cause. This issue has been rumbling on for months, at least a handful of invited users had not even heard of the problem.

Users need to be aware that 600 cycle kits are most likely not to work for long amplicons as the ends of both reads will drop in quality - badly.

The sentiment at the meeting was that issuing a quality notification to make sure customers are aware of this risks damaging confidence in these kits. Whilst many users will be unaffected (small random fragment genomes) or anyone sequencing under 500bp, the kit is clearly not delivering as expected in many long-read situations. Shareholders and investment banks may react negatively to a PQN, but without it too many users are in the dark.

Speak to your FAS and Sales team.

[pmiguel]

It sounds to me like a diversity issue, since the lowest diversity will be near the beginning. Are you using staggered primers, or spiking in anything with more diversity? There's also cluster density; even if you follow the same protocol, a higher cluster density will suffer more from a lack of diversity.

Brian,
Diversity should no longer be an issue with MiSeq runs. Or HiSeq runs for that matter. Modern versions of the Control Software for either instrument recognize low bias sequence and correct for it.

You just need enough phiX library (or other high diversity library) to put a few clusters per image (5-10% is fine) to keep the focus on track.

That said, and even though I've seen plenty of zero diversity MiSeq runs that look just fine, I'm suspicious at this point that the instrument might be thrown off by low-but-not-single-plex diversity. Since I've seen some poor results from sample with 4-6 amplicons. Or samples that start off single-plex but then become diverse.

But Illumina replaced the reagents for those runs. We just need to get new library pools from the customers to try again. (These were 500 cycle reagents and had a different failure mode -- good sequence for about 100 bases of each read, followed by terrible sequence quality.)

--
Phillip

[Brian Bushnell]

It sounds to me like a diversity issue, since the lowest diversity will be near the beginning. Are you using staggered primers, or spiking in anything with more diversity? There's also cluster density; even if you follow the same protocol, a higher cluster density will suffer more from a lack of diversity.

[toneta2013]

This happens across all runs (4 up to now), so I assume it's not a bouble. We're running the standard 16S metagenome protocol from Illumina, so I guess our diversity shouldn't be much different from anyone else's running this protocol?

[GenoMax]

We're struggling with the reagent issue in Norway as well. You've all been describing the early loss of good quality data, however we also see a quality dip in R1 around bp 6 to 20. Anyone familiar with this? I've attach a file showing a typical fastQC output on our 16S metagenom data.

If this was a one time observation it could be because of a glitch with the sequencer (e.g. small bubble in liquid path). If it is reproducible across runs, you may have a product that has low nucleotide diversity around that region seen in R1. Comments?

[toneta2013]

We're struggling with the reagent issue in Norway as well. You've all been describing the early loss of good quality data, however we also see a quality dip in R1 around bp 6 to 20. Anyone familiar with this? I've attach a file showing a typical fastQC output on our 16S metagenom data.

[martinjf]

same issue here, our last 8 kits V3 2x300 runs were reimbursed (well they sent new kits but the problem is not solved....
We rechecked all runs and the problem appeared last february at least for us.
We checked three different miseq instruments, same issue.
Illumina recognizes here too that they have an issue with the reagents, no precise lot is isolated so far, even the most recent ones fail.
Around us most providers suspended their operations with V3 2x300 for now.

hopefully they will resolve this issue soon,

have faith

[kmcarr]

Given the amount of trouble they are having with this, I think they should have separate specs for low-plex (eg, single-plex) amplicon runs in their sales advertisements. That way people know they are not going to be able to push the system as much when doing amplicon work.

This point deserves repeating Phillip. Managing client expectations during initial consultations for MiSeq amplicon projects is huge. I generally advise them to plan their experiments with the expectation of 9-10 million usable read pairs from a MiSeq v2 PE250 run for amplicon libraries. I explain that for amplicon libraries we lower the cluster density; that the PhiX control library will take up 5-10% of the reads and with large, complex index sets the fraction of reads with non-assignable indexes is higher.

[pmiguel]

Modern MiSeq Control Software doesn't seem to benefit from higher phiX content -- beyond ~5-10% necessary for the instrument focus on. But dropping the clustering concentration down enough does seem to help. Obviously you can then end up with runs that are below Illumina's specifications for a given MiSeq run.

Given the amount of trouble they are having with this, I think they should have separate specs for low-plex (eg, single-plex) amplicon runs in their sales advertisements. That way people know they are not going to be able to push the system as much when doing amplicon work.

--
Phillip

[Brian Bushnell]

You can calculate the insert size distribution with BBMerge like this:

bbmerge.sh in=r1.fq in2=r2.fq ihist=ihist.txt vloose qtrim2=r

(those settings are designed for low-quality reads).

This will give you an indication of what fraction of your inserts are shorter than read length, and thus whether the third bullet point is a plausible explanation.

Also, can you post a base frequency histogram (by position)? And, how much PhiX are you using; are you multiplexing different amplicons together, or is there just one kind in a lane; and are you using variably-sized sequencing primers to stagger the read starts?

[petersm3]

Low %>Q30 on R2 for 2x300 libraries

We have had numerous quality complaints from our customers performing 2x300 and 2x250 MiSeq runs (predominately amplicon libraries) where the %>Q30 is approximately 60% or less due to the “poor” quality of Read 2, and in some instances a significant number of clusters do not pass filter.

In a handful of cases, Illumina tech support cited issues after reviewing SAV, e.g.,

• "There may have been a transient fluidics issue that occurred at the beginning of the run"
• "Loss of focus at the end of the run likely caused by over clustering"
• "Quality drop corresponds with an increase of A base calling. This is indicative of running out of insert to sequence."

Illumina replaced several kits and dispatched a field technician to adjust the MiSeq due to a non-optimal Z-focus setting. Even after this adjustment, we are still seeing low PF rates for our 2x300 and 2x250 runs, and tech support states that our MiSeq is operating within spec. The only suggestions offered by Illumina are that the samples run, e.g., amplicon libraries, are low in base diversity and that we should continue to decrease loading concentration and increase PhiX spike-in, which we already do and does not appear to significantly improve the situation.

[punkka]

We are having the same issues with V3, quality dropping in the forward read even before first 80-100 bases. We tried to get some more information as it seems it is not related to any batch. Has anyone any idea what might be the problem? In our lab we are running lots of targeted DNA sequencing, even using same library preps and MASTRs and we do see decline in quality since the middle of this summer and it affects different LOTs. As a response we get replacement kits, but yeah that does not solve the issue of wasting time and producing bad data.

[markbvdh]

The last few Miseq 600 cycles kits we've used have all had rapid increasing error rates after 100-150 cycles of the forward and reverse reads. The FAS's here (Australia) have been alluding to a (global) problem for quite a few months now with the 500 & 600 cycle Miseq kits, but there's been no formal batch flagging from Illumina HQ to date.... Due to the time everyone wastes in re-runs & troubleshooting, in my opinion, this should have happened by now. Sure, they offer to replace kits - but what about the time spent?

Mark