Loss of tiles with v3 chemistry

[hoisinjl]

Ever since we have switched to v3 chemistry, we have noticed that we are loosing tiles randomly during imaging. There seem to be 3 specific phenotypes:

1. Loosing a few tiles for a cycle, and then it recovers. This can happen only once during the runs or several times.
2. Loosing the same tile many tiles during the run, causing the tile not to be analyzed at all, reducing the pass filter rate.
3. Loosing an entire surface (top or bottom) for a cycle.

Has anyone else seen this? The bottom, middle surface (8 tiles) seems to be the most common to be lost, but have seen it in other tiles.

[kmcarr]

We have just recently had our HiSeq installed and during the test run I noted a number of phenotype 3 occurrences. They were always the bottom surface, swath #2 (middle); they occurred in several different lanes (6 of the 8) and from cycles throughout the run (it was a 2x100 run). In all there were 25 of these bad scans over course of our test run. I haven't been able to talk to our FAS yet about this to determine if this frequency is normal.

During the run our FAS explained that this can occur if there is a bubble near the inlet port of the lane. This causes a focus error at the beginning of the scan so the entire scan for that swath will be out of focus. It makes sense that this would be more likely for scans of the bottom surface since it would be trying to focus through the bubble. I'm not sure why they were uniformly in the middle swath.

I'm worried about the frequency we observed. Considering that a bad swath blows 16.7% of the clusters in that lane for that cycle that's huge. Have that happen a couple of times in the same lane/swath and before you know it you've lost 1/6th of your reads from that lane for having too many N's.

[GenoMax]

We have several HiSeq's and have very rarely had problems with dropped tiles. It sounds like a visit from an illumina engineer may be needed to check your HiSeq.

[pmiguel]

We saw dropped tiles of this sort in our HiScanSQ prior to complete blockages in some of our lines solved by a few service calls to iron out fluidics issues. I have heard these are endemic to new HiSeqs and HiScanSQs, but not old ones. Since then I mainly see dropped tiles in bar code cycles where low numbers of indexes are in the lane.

Strangely, some of the dropped tiles did not seem to have focus problems -- the clusters were quite clear.

--
Phillip

[GW_OK]

Well I've been on the phone with tech support about this for two runs now. There's an actual name for the phenotype: "BMS" (Bottom Middle Swath). You can call up tech support and say "I've got BMS" and they'll know what you're talking about. I've been informed that Illumina has their best people on it, but as long as you get "enough" data it's going to be a shoulder-shrugging issue.

Reminds me a little of the v1 hiseq flowcells...

[kmcarr]

Well I've been on the phone with tech support about this for two runs now. There's an actual name for the phenotype: "BMS" (Bottom Middle Swath). You can call up tech support and say "I've got BMS" and they'll know what you're talking about. I've been informed that Illumina has their best people on it, but as long as you get "enough" data it's going to be a shoulder-shrugging issue.

Reminds me a little of the v1 hiseq flowcells...

Thanks for the report GW_OK. It's nice to know we're not the only ones.

[hoisinjl]

We have figured out that for a tile not to be completed analyzed, you have to loose focus twice in the first 24 tiles (i.e. before the chastity filter). We just had the y-axis on our HiSeqs adjusted so that the imaging is more towards the back of the flowcell. We are monitoring the next couple of runs closely to see if we see a decrease in the number of occurrences. I will update when I know more.

[josdegraaf]

Same here, is indeed a known problem due to bubbles in the lanes during scanning. Maybe due to bigger lanes?

[sehrrot]

Yes it's due to the physically wider lane, which introduce more bubbles. When I first ran V3 fc, I changed gaskets 10 times. But still substantial bubbles. We also found some phenotype issues (spikes) but recovered after few cycles (but luckily we didn't lose any data and got over 300Gb output).

This issue only happens in bottom middle surface, as Illumina admits. Hope it will be fixed soon. BTW at this moment we can do is only removing bubbles as much as we possible. Also, for the first time V3 user, do 1 maintenence wash + 2-3 times water washes + double priming before you start actual run. It will remove and make V3 environment in your Hiseq syringes (remove all V2 reagents in there)