How to avoid cluster density gradient across tiles?

[pmiguel]

In our last couple of runs on our HiScanSQ (which uses HiSeq chemistry/flowcells) we get a cluster density picture that looks like:



(I masked out 3 of the lanes because they had high primer-dimer contamination that threw off the cluster calling.)

Do any of you typically see this much cluster density variation from top to bottom of a lane (around 2x for this run.) More importantly, does anyone know how to keep the gradient to a minimum?

--
Phillip

[pmiguel]

Wow, not one answer. Guess we are the only ones?
Anyway here are some more symptoms:
Our cBot seemed to need its coolant level topped up frequently. Our most recent run had a similar cluster density issue that I mention here:

http://seqanswers.com/forums/showthread.php?t=13121

The flow cell seemed "wet" when being transferred to the sequencer and had to be cleaned. Also we got a mysterious "temperature out of range" message from the cBot. (I can post the completely useless graph of the run temps the cBot displays, if anyone cares.)

Have you figured it out?

Okay, the Illumina Field Service Engineer came and opened up the cBot -- sure enough the temperature block was leaking coolant. Not enough to be visible without opening up the instrument, but enough to create a temperature gradient across the flow cell, evidently.

--
Phillip

[sehrrot]

Hi Phillip

We've seen this all the time - just guess the gradient would be simply formed in a way of flow: templates are injected from the bottom swarths in that picture. This is our best guess. No solution for this, so far but one thing you may try is making balance of your cBot, slanted cBot installation might worsen this kind of the gradient issue.

Regarding 'temperature issue', have you ever updated your cBot's software? I'd seen that notice until I updated cBot's new software.

[niceday]

I have seen gradients but not to this extent.
It looks as if the gradient is from cluster formation because it wouldnt be so consistent for each cycle during sequencing.
Have you had your cbot serviced because it looks like there might be a slower flow or reduced flow at a guess.

[sehrrot]

Hi pmiguel.

Since I replied you and before then, I've oberved that issue. Now we fixed tilting cBot and soon or later new run will start. I can see whether it will be helpful or not..

[pmiguel]

I should have updated this thread to say that replacement of the cracked heating block did not completely fix the gradient. That is we still saw substantial (>2x difference in raw cluster density from top to bottom tiles) gradients.

The engineer came out and checked the calibration of the block. There was no temperature difference between the ends of the block. However the high temp set point was off by a few degrees. The low temp set point was fine. It seemed like this was unlikely to be the issue. But a subsequent run did not show a substantial gradient.

--
Phillip

[cnicolet]

Cluster gradient

We routinely see similar gradients, and see it regardless of the cBot we use (we have two). I think this is just a reflection of reagent flow, as noted in an earlier thread. The cluster station gave similar gradients. It wouldn't be that big of a deal, but lately I have been hearing the Illumna tech support people really hammer on cluster number. If you have an issue and your cluster density varies at all from what they consider perfect, they'll say it's that. The problem with the gradient is you never get the perfect number except at the one section of the flow cell.

[pmiguel]

We routinely see similar gradients, and see it regardless of the cBot we use (we have two). I think this is just a reflection of reagent flow, as noted in an earlier thread. The cluster station gave similar gradients. It wouldn't be that big of a deal, but lately I have been hearing the Illumna tech support people really hammer on cluster number. If you have an issue and your cluster density varies at all from what they consider perfect, they'll say it's that. The problem with the gradient is you never get the perfect number except at the one section of the flow cell.

After replacing the block in the cBot that was leaking fluid and re-calibrating its upper set point, our gradient became much less steep:


Above, are data from 3 lanes with average raw cluster density at, or very close to the maximum recommended 850 K/mm2. Note that the maximum lane gradient is only 1.34x for lane 6.



For the first 100 cycles (the forward read) quality performance is similar all the way up to nearly 970 K/mm2.



Although density does appear to play a role in a (few) poor performing tiles in later cycles of the second read. I should add: the really low points there all derive from the final (non-phasing corrected) cycle 209

(This was a 101+7+101 cycle R1+index read + R2 (PE) run on a HiScanSQ.)

Anyway, the density gradient is Illumina's responsibility, so if you are seeing high cluster density gradients and are under a service contract, then you should demand that Illumina fix the issue. Within reason. I think anything under 1.5x would be reasonable. If you are frequently seeing 2x differences, I think there is an issue.

--
Phillip

[cnicolet]

Thanks Phillip, this is very helpful. I have been getting good advice from you since my days running a Sanger sequencing core!

[TonyBrooks]

We've just seen something similar on our latest HiSeq run.



I checked the cBot coolant and it seems fine. I'm going to contact Illumina.

[pmiguel]

Our coolant issue was that the block itself was cracked and leaking coolant. Our cBot's manifold movement had to be recalibrated as well before the problem subsided.

"Subsided", not disappeared. There is a certain amount of cluster density variation from top to bottom. It is variable though, and I don't know what causes it to be higher in one case than in another.

--
Phillip