Illumina MiSeq-run: Bad quality per tile. How should i interpret this fastqc-image?

[jov14]

I recently got very bad sequencing results from a Miseq Illumina run. The average Sequence qualities declined very early, leading to low read lengths after quality clipping.

When I looked at the respective FastQC-results i noticed a very strange pattern in the "Quality per tile"-Plot (see attached .png file):

I get less than average quality for cycles 80-280 almost exclusively at tiles 1112-1119 and then again at tiles 2115-2119.
What I find extremely strange is the apparent "smear" of bad quality from tiles 1112-1119, which then seems to stop abruptly. I do not quite understand what is up with tiles 1120-2100: Are they simply not shown in this plot, or should I assume that Tiles 1119 and 2101 are neighbors?

usually when I see examples of bad quality-per-tile patterns, bad quality tiles seem to be more distributed along the flow cell and have consistently bad quality during all cycles (e.g. this example).
In my case, however i seem to have a double smear-pattern of temporarily bad quality tiles, broadly around cycle 175.

Does anybody have experience in interpreting such plots?
I would like to understand what the tile-designations "1101"-"2119" mean, exactly.
Could something like bubbles have formed themselves at cycle ~80 and then just burst sometime around cycle 280?
In that case, how can it be that this pattern seems to occur twice on the flow cell around roughly the same time? Pure coincidence?

[dpryan]

Those tiles are on opposite sides of the flow cell (the first number in the tile denotes whether it's on the top or bottom). So I suppose it could be a bubble, since that might affect the camera focusing on the other side. Having said that, the core facility and Illumina can probably give you a better answer, since they deal with the machines daily.

[jov14]

Thanks for the reply. So as i understand it, the FASTQC-plot shows two opposide sides of my flow cell in part only, in my case?

I'll also try the illumina customer support. However, at least from my previous experience, you often do not get much info on what actually caused any reported problem and/or why. Often you just simply get a replacement for the reagents if Illumina thinks there might be a problem with the reagents or the machine, in the hope that it doesn't occur again next time.

Therfore I thought I'd ask the community first for detailed infos suggestions. Not sure if i'm looking at a problem caused by (e.g.) a faulty flow cell or by something that could have been prevented during sample preperation/loading.

[pmiguel]

Thanks for the reply. So as i understand it, the FASTQC-plot shows two opposide sides of my flow cell in part only, in my case?

The tile numbers are of a format:

XYZZ

Where "X" is the surface of the flowcell -- 1= top, 2=bottom;
"Y" is the swath number. MiSeqs only have a single swath, so you can ignore this position;
"ZZ" is the tile -- should be numbered "01" at one end of the flowcell, and "19" at the other end (for a v2 MiSeq flowcell).

So it looks like your issue occurs mid-run, impacting one end of your flowcell. Both surfaces are affected, but the upper surface, more than the bottom.

I have no idea what would cause this. A bubble in the flowcell during scanning should cause more issues for the bottom surface as there will be issues focusing through the bubble.

How do the thumbnail images of the tiles look during the low quality cycles?

--
Phillip