How to reduce flowcell bubbles in the HiSeq?

[pmiguel]

A maddening property of the HiSeq is that it does not seem to take any steps to eliminate bubbles from the liquids it passes through the flow cell. Does anyone have methods to reduce their numbers?

Generally this has no effect on the top surface tiles but bubbles in the scan buffer prevent clusters beneath them from being scanned. Worse, bubbles near the bottom ports of the flow cell can cause imaging issues for the entire swath! (This is called "bottom middle swath" or "BMS".)

Further, the number of bubbles seems highest in the first few cycles, the very worst time for them. Any clusters covered by a bubble during the first 4 cycles will not pass filter -- are effectively lost. BMS causes 1/6th of the clusters in a lane to be lost.

Just to be clear, I do not mean bubbles being pulled through the rims of the flow cell gaskets ports. I can see these during a pump test using a pair of +6 "reading" glasses I bought from Amazon and a flashlight (or "torch" as it is known in other "english" speaking countries). This is always caused by tiny fragments of dust laying across the rim of the effected gasket port. Again, keen eyesight or +6 glasses may be needed to visualize these.

I mean the bubbles that are either in the scan buffer or form during pumping. (I presume the pumps produce enough vacuum to cause a certain number of them to form by solution de-gassing.)

--
Phillip

[josdegraaf]

Illumina recommended us to do a maintenance wash before the run using the same gaskets as during the real run..

[pmiguel]

Illumina recommended us to do a maintenance wash before the run using the same gaskets as during the real run..

How does that work for you?

By the way, I follow exactly that "gasket change/maintenance wash" protocol. I am not sure what problem that is trying to solve -- maybe it removes any solvents that might leach from the gaskets?

However it does not solve the issue I am talking about here. These are not bubbles coming from the gasket seal, they are present in the lines leading to the seals. You can see the bubbles coming through the tubes drilled through the transparent amber plastic blocks leading to the seals. Although you might need a magnifying glass and a light source to see them.

I think these bubbles are from gas dissolved in the scanning buffer coming out of solution during pumping. That is my guess anyway.

--
Phillip

[GenoMax]

Are there temperature fluctuations (day/night cycles or otherwise) in the room where your HiSeq resides?

I remember bubbles being a issue in the past when we had problems with temperature control (though I can't definitely say they were related, just anecdotal thinking). Lately I have not heard of bubbles being a problem.

[pmiguel]

Are there temperature fluctuations (day/night cycles or otherwise) in the room where your HiSeq resides?

I remember bubbles being a issue in the past when we had problems with temperature control (though I can't definitely say they were related, just anecdotal thinking). Lately I have not heard of bubbles being a problem.

I guess there would be some, but nothing much. We are in a sub-basement lab and I keep it pretty cool. Usually between 19-21 oC.

Are you sure they are not an issue for you? My assay is here is the error rate for the top surface of a flowcell:



Here it is for the bottom surface of that same flowcell:



If I go to one of the cycles with a high error rate tile, look at that cycle in the flowcell chart, I can see that tile, click on it and SAV pops over to the imaging tab and I see:



Pretty much every time.

You don't see those?

--
Phillip

[GenoMax]

I had a look at a couple of the recent HiSeq (2000 and 2500) flowcells (following your path above). I checked a few cycles and high(er) error tiles. Did not see any bubbles on tiles I checked.

I am not on experimental side of things but keep an ear open for problems like this since they ultimately affect data.

[pmiguel]

I had a look at a couple of the recent HiSeq (2000 and 2500) flowcells (following your path above). I checked a few cycles and high(er) error tiles. Did not see any bubbles on tiles I checked.

I am not on experimental side of things but keep an ear open for problems like this since they ultimately affect data.

Did you see the big difference between the top and bottom surface on the error rate plot?

We only have one instrument, so I don't know if it is normal to have that number of bubbles.

--
Phillip

[GenoMax]

Did you see the big difference between the top and bottom surface on the error rate plot?

We only have one instrument, so I don't know if it is normal to have that number of bubbles.

--
Phillip

The difference was not as dramatic as your plots but the top surface was better. We have several HiSeq's and I have not heard of recurring bubble issues.

You may want to ask your friendly illumina engineer to look at the possibility of a small leak somewhere, specially if you see the bubbles regularly.

[josdegraaf]

worked great for us, before we had the same as you show here. After we start doing maintenance before every run (using the gaskets for the run) and keep washing with water till the run starts we got rid of it. I do occasionally see some bubbles but marginal.

[pmiguel]

The difference was not as dramatic as your plots but the top surface was better. We have several HiSeq's and I have not heard of recurring bubble issues.

I can only speculate, but if this is a common issue, then I don't think it would be noticed for the most part. You actually need to make the link between "some high error tiles in the bottom surface of my flowcell" and "transient bubbles in my flowcell".
Because they are transient and not that common, there is just a slight degradation in the data quality from reads from the bottom surface. These end up being "N's" or very low quality bases in the final sequence.

You may want to ask your friendly illumina engineer to look at the possibility of a small leak somewhere, specially if you see the bubbles regularly.

I do, but there are a certain number of bubbles you see in flow cell lanes during, for example, a pump check. They actually train you to see these small bubbles coming through as a method to ensure there is flow.

To me that says some bubbles are "situation normal". Given that, then you expect a number of bubbles in the scan buffer as well.

At the risk of testing you patience on this, the error plot I showed previously:



If you right-click on that pane in SAV and choose "Autoscale" you get a full scale assessment of the bad tiles:



So maybe it isn't the tiles that have 4% error rates that have the bubbles, but the tiles that have higher ones? The flowcell pane of SAV appears to cap the error rate color scale at 4% unless you manually adjust it higher. I have to hover the cursor over the scale and use my scroll wheel to do that. Then I can focus on just the tiles that have pretty high (>4%) error rates:



Then you can click and drag on the color scale to get to a place in their palate that really highlights the high error tiles:



So I can see nine high error tiles:


For each the cause would be the bubble that is clearly visible.
You don't see that at all?
How about anyone else?
If not, then my instrument has an issue, that could presumably be fixed.

I guess on the scale of things, the issue is minor. It only effects the clusters covered for that one cycle. 9 of 384 tiles or 2.3%. Even then the tiles only suffer a few percent loss in data. So maybe I am being too "glass half empty here"?
--
Phillip

[GW_OK]

I can report we see it here as well, on both of our Hiseqs. I've tried degassing the scan mix in a vacuum chamber but that didn't do anything. We've also tried the maintenance wash fix, but that didn't solve it either.

[GenoMax]

Phillip,

I followed your screenshots with a couple different flowcells from 2 different HiSeq. I am not seeing any bubbles in the images for tiles that have high error rates.

What about the very high error rate tiles (65%) around cycle 60 and 75 in your screen shots. What is going on there?

[pmiguel]

Phillip,

I followed your screenshots with a couple different flowcells from 2 different HiSeq. I am not seeing any bubbles in the images for tiles that have high error rates.

What about the very high error rate tiles (65%) around cycle 60 and 75 in your screen shots. What is going on there?

Bottom middle swath.
(Again presumably caused by a bubble -- but at a critical point in the flow swath that caused the instrument to mis-focus for that swath...)

[pmiguel]

I can report we see it here as well, on both of our Hiseqs. I've tried degassing the scan mix in a vacuum chamber but that didn't do anything. We've also tried the maintenance wash fix, but that didn't solve it either.

Thanks. At least now I know it isn't just our instrument.

I guess what is mystifying is that GenoMax does not see this...


--
Phillip

[GenoMax]

I can report we see it here as well, on both of our Hiseqs. I've tried degassing the scan mix in a vacuum chamber but that didn't do anything. We've also tried the maintenance wash fix, but that didn't solve it either.

Do you see this on every flowcell? All the way down to the images like Phillip posted above?

Thanks. At least now I know it isn't just our instrument.

I guess what is mystifying is that GenoMax does not see this...

The bubbles were not there in the images from the two flowcells I looked at. I will check with the lab folks here to see if they have anything to add.