Low Cluster PF % and highly skewed Base %

[sequencer13]

Hi,

My run (2x150 PE, MiSeq) initially looked good, with the pretty much the optimum cluster density for our machine. Unfortunately the cluster PF % is down at 52 %, which is clearly a significant issue.

In addition the base % varies wildly:

C ~65%
A ~30%
T ~2.5%
G ~2.5%

The Q score is poor with a low >30 %.

Has anybody seen anything like this before, or can offer any advice as to what may have gone wrong?

Many thanks for your consideration.

[GenoMax]

Is this run likely to have either of the following characteristics?

low nucleotide diversity and/or short(er) than expected inserts.

Are you using the latest version of MCS/RTA available from illumina? Can you post a screen grab of the heat-map of the scores?

[HESmith]

Not sure about the MiSeq, but our HiSeq uses two lasers for excitation: one for A/C, the other for G/T. A laser malfunction would produce the results you see.

[sequencer13]

Hi - thanks for your reply. I would not expect the library to be very low diversity, and the insert size should be approx. 300.

The Q scores are abysmal, picture attached - the cause could be the Phas/Prephas (%) of 0.503/0.549 for read 1. It will be interesting to see if the Phas/Prephas (%) is similar or normal for read 3.

G/T laser failure perhaps?

Edit: HESmith - it is the same on the MiSeq and we were thinking this seems plausible. Have to wait for the run to finish to check though >.<

[GenoMax]

There may be some problem with this library to begin with (unless it has been run before and know to be good) since the Q-scores are bad.

Since MiSeq is lot more integrated with the control software I assume it will throw up an error (the dreaded red flashing warning) if a laser was not working right.

[sequencer13]

Valid point, no red light of death here.

It's possible these particular libraries (three technical replicates) are bad, though they are fairly similar to previous work that was fine.

The overall profile (example attached) looked decent, but I guess it is possible something went wrong during library prep / other manipulations.

[GW_OK]

Could you post the base intensity plots for the run?

[sequencer13]

Yep, is this the plot you meant?

Edit: As an update on a previous point, current Phas/Prephas is 0.012 / 0.055 for read 3

[GW_OK]

Based off the intensity plots it doesn't look like a low diversity or short library. It also doesn't look like a laser's faulty as the intensity curves look like what I usually see on my Miseq.
You've got a real head-scratcher here. Might have to see what the fastq's look like.

[sequencer13]

Yep we agree with that.

Another 'interesting' phenomenon is that the base % has not changed after turn around. C is still at 70 %...which is physically impossible...?

Edit: C is off the scale, can't seem to adjust the axis range on the remote SAV.

[GenoMax]

BTW: What is the exact cluster density being reported? It is also likely that this run may be overloaded. In certain instances MiSeq will put a number up that does not translate to real cluster numbers.

[sequencer13]

The cluster density is ~1400k +-5

(You may well think this very high, but this is around where we tend to work for good results)

Edit: you may have a point with your edit, if this (high) density is an under estimate it would cause problems.

[GenoMax]

The cluster density is ~1400k +-5

(You may well think this very high, but this is around where we tend to work for good results)

Edit: you may have a point with your edit, if this (high) density is an under estimate it would cause problems.

I would suggest looking at the images (though that also is not always a reliable metric either). MiSeq will keep working beyond the "design limit" reasonably well but once over the "cliff" the fall (in Q values) will be pretty drastic.

[sequencer13]

I've had a quick scan through some images and none flag up any obvious issues to my eye. The C channel does look somewhat more intense but even so a,t,g,c images for a couple difference cycles look ok. Are there any specific issues you would advise to look out for?

Also, thanks to everybody who has been helping out with this - it's much appreciated!

[mcnelson.phd]

So, we can rule out any sort of "laser failure" because the MiSeq uses LEDs instead of an actual laser for exciting the flours. LEDs can fail, but they're pretty stable devices so I'd say it's unlikely to be the cause here. EDIT: And as Geno noted earlier, most likely an LED failure would give you the Red X of Doom warning and kill the run before it even began.

More likely, you're so over clustered that MCS/RTA can't properly resolve the clusters, so signals are bleeding all over the place. We've seen similar but slightly different results with high-GC genome runs that get over clustered. At some point, MCS gives up on trying to accurately report cluster density and simply give a big number.

For your % base skew, from my experience RTA will assign A bases for clusters where the signal is too poor to accurately determine the correct base. This generally occurs after you've sequenced through the entire molecule and into the anchoring oligo.

Best thing to do now is kill your run if it's still going and re-run the samples at a saner cluster density like 1000K. Illumina still only advises you to shoot for 850K for most runs, even though they target 1250K for a phiX control run. If you're pushing well beyond that, then you're on your own really.