2x300 MiSeq read 2 failures due to reagent?

[ksawatzki]

I haven't taken a detailed look at the fastq since it's a complete loss for us (even though R1 is completely normal, we require both) and also because we're not involved with the sequencers themselves. However, on cursory glance, it looks like the dominant two read types are poly-C runs and N-dominated. Even reads with some nucleotide diversity pretty much are universally 0 scores. The sequences definitely do not make sense and are obviously not biological. Just to clarify, all the R2 reads are the full 300bp long.

[GenoMax]

@dfhdfh
I'm not sure what the range of the problem is, but I've attached the Prinseq QC output for Read2 of one of my reagent-based failed runs. It just... immediately stops working.

Can you please clarify if there is no sequence beyond that point (N's?) or just that the Q-scores effectively become zero? Does the sequence make sense (if it is present, even if the Q-scores are all 0)?

[ksawatzki]

@dfhdfh
I'm not sure what the range of the problem is, but I've attached the Prinseq QC output for Read2 of one of my reagent-based failed runs. It just... immediately stops working.

[dfhdfh]

Hm, that's concerning. I'm running a 2x300 v3 run at the moment, and now, after ~500 cycles, I have a sudden drop in quality. It's not completely finished, yet, but cycles 500-580 have a quality of <10. Up until cycle 500 everything was fine with >70 % >Q30. Now it has dropped to 60 % and is still dwindling. I will contact Illumina tomorrow when the run has finished.

Oh, by the way: Illumina RTA stopped working sometime during the run. I didn't close the program, however, because everything was still running. RTA calls the bases and quality during the run, so I'm not sure how I'm still getting data?

[pmiguel]

We did a 600 cycle run about a week ago. Different failure mode. Read 1 prephasing was sky-high but phasing was low. (Read 1 0.062/0.974; Read 4 0.299/1.127) After 80 bases quality dropped and the error rate rocketed to 70%. Tech support had us do a volume test, etc. All passed. Obviously we asked that Illumina check the lot # we were using. But they said there was no issue with that lot. And maybe it was just the cassette that was bad.

So another run was started, same instrument. Pretty much the same thing happened. (Read 1 0.080/0.475 Read 4 0.349/0.667 phas/prephas %)

The libraries were customer created amplicons -- but a mixture of ~9 loci. So while there was library bias, it was minor compared to the single-plex amplicons we often run.

After the 2nd failure Illumina instrument tech support wanted a phiX 50 cycle run done on the instrument and Illumina library tech support to contact us. I asked if they thought there were library issues that could cause extremely high pre-phasing. Also the phiX and 50 cycle cassette didn't ship because none were not available that day. At which point they told us they were sending out an engineer. He arrived the next day, found some components he didn't like in the instrument and replaced them.

Meanwhile we repeated the run on our 2nd MiSeq using replacement reagents. This run looked as good as any v3 run we've ever done. So the libraries themselves were not the issue.

However, one interesting tidbit of information: tech support told us:

In general for sequencing runs on all Illumina platforms, library composition can have an effect on the reported phasing/prephasing metrics. When the base diversity is not well balanced, it can cause issues for the software to accurately estimate the phasing/matrix information. When sequencing these types of libraries, the raw cluster density should be between 500k/mm2-1000 or 1100k/mm2 for the software to accurately generate a cluster map and extract metrics.

I'd never heard of a lower end issue -- unless you were really low, like below 200. I don't actually think this had anything to do with our issue, but I can't rule it out. I had noticed the cluster densities were low on the first 2 runs. (477 and 305 for run1 and run2). Then I noticed the person doing the runs had an error in their concentration calculations. So the 3rd run was 772.

In situations like this, my tendency is to suspect what I call "multiple systems failure". Aka "When it rains, it pours." That is, most stable systems have a robustness that tolerates 1 or 2 minor issues without substantially degrading the quality of data being collected. So when you do start seeing issues, they don't usually have a single cause, they are the result of several factors falling below optimum.

Here are the culprits I would tend to suspect for this issue:
(1) Aging MiSeq components.
(2) Cleaning/maintenance protocols for MiSeq insufficient. Possibly resulting in biofilms in some percentage of instruments.
(3) Systemic issues with v3 MiSeq reagents causing them to perform slightly below spec most of the time and substantially below spec on occasion.
(4) And the usual reporting/confirmation bias. Few people jump on a forum like SeqAnswers to say how smoothly their MiSeq runs are going.

--
Phillip

[GenoMax]

@Bukowski: Do you know that this is a widespread problem that affects all MiSeq reagents produced during some time period (do you have the lot # that were affected?). I don't remember seeing anything from Illumina Support in the last couple of months.

[ksawatzki]

Great, that's very reassuring to hear. I just couldn't find anything that confirmed there was a known problem, the core tells me that they have it handled. Thanks!

[Bukowski]

Yes this is a reagent issue. Illumina have acknowledged this. Contact Illumina. You should receive replacement reagents.

[ksawatzki]

We have been running many libraries with a core using the PE 300 kit on their MiSeqs. We've been working with them successfully for more than a year, and I've been using a set protocol for at least a couple successful runs.

Two months ago, they ran three of our libraries and read 2 (read 3, technically, we have indices) was completely unusable. Quality was well below 20 within just a few cycles. Read 1 was otherwise normal. Unfortunately for our project, our required reads are >300bp so we require both reads. I was immediately concerned that something was wrong with the adapters or index mixes, but they informed us that they saw the same problem among multiple samples. They said that it was a problem with Illumina's reagents. I have no reason to doubt them, but I didn't see anything here on a cursory search and they still haven't sequenced anything.

Does anyone know about this? What happened and when is it expected to be fixed? We're in a huge time crunch as our money source is about to cut us off and I have a ton of libraries that need sequencing. Any information and help would be appreciated.