Tuesday, 27 November 2012

Don't count your clusters before they hatch...

I've dropped a few veiled hints about my previous disastrous runs on the MiSeq (which I do plan to write up eventually, but the anguish is still too fresh), but today we got samples from our revised strategy running on the machine.

This is the first of our samples we've run ourselves (well, that we set up at the genomics expert's direction), and as such we were pretty excited to watch the data start to roll out in real time, particularly in light of our previous failures.

Then it happened. The first data to pop up in BaseSpace looked like this:

There's a couple of things that put the fear in me here.

The first is the upper panel; not only are the intensities quite low, some even below 200, they are going up.

The second is the more worrying; from our lower panel we can see that not only do we have relatively low cluster density (300k/mm2 when we aim for 800k/mm2, see blue box), but it looks like none of them have passed filter (indicated by the green box smooshed into zero on the axis).

Like I say, this is only the first run I've watched from the very beginning, so I was hoping that this was just a feature of it counting the clusters in earlier cycles before finishing the filter checks at cycle 25, as I'd read in some Illumina documentation.

However, I wasn't sure, so I spent a paranoid five minutes trying to double check whether this was the case, to no avail at all (and obviously I couldn't find the Illumina reference where I'd read it in the first place).

So it was with much relief that, when cycle 25 ticked over, this happened:

Phew. We might not have many clusters, but at least most of them are passing filter.

To anyone who's done more than one run this post probably seems like a lot of fuss about nothing, but I certainly spent those five minutes googling with my stomach turning over. Hopefully if anyone else finds themselves in a similar scenario, they might find their way here and heed this advice; just wait for cycle 25.

As an aside, the run is currently at cycle 150, and the intensities are still rising. Is this perhaps a feature of the v2 kits, or something more anomalous? Until I have a play with the sequence tomorrow, I'm reluctant to speculate on what this might mean.


The machine's still running, but I've got some feedback about the funny error profile, after I posted it on twitter. It turns out that this is to be expected in v2 kits. Good to know!

No comments:

Post a Comment