Illumina NextSeq non homogenous cluster density

[thiNGS]

Hello. I am performing whole genome sequencing of bacterial strains and I got very variable cluster density (and hence signal intensity) over the four lanes of the flow cell. Has anyone ever seen this and can help me troubleshooting? Thanks for any suggestions you might want to give me.

[JoeKutch]

Intensity =/= cluster density, though they are related. There can be variations in intensity from imperfections in the flow cell, smudges on the surface, and focusing. Even reagent age/batch.

The first image I've attached is from a low density (150K/mm^2) and the second is from a high density (280K/mm^2). Notice that the lower density run has a much higher intensity (see the scale on the right). Despite the variations in intensity, all four lanes from the first run were within 5% (<3million reads) of each other in terms of reads PF. A similar percentage variability was seen in the high density, lower intensity run.

As long as there are no dramatic quality drops associated with the areas of low intensity, I wouldn't be concerned.

[thiNGS]

Many thanks for your reply. I see this pattern with all my runs despite reagents and flow cell batches though. I wonder if my library concentration might be too high and so if I am "overloading" the flow cell. The problem is in lanes like 3 and 4 I lose quite a lot of reads because at least 15% won't pass the filter. Do you think overloading the flow cell might be a possible reason?

[JoeKutch]

Without knowing the cluster densities reported by your instrument, I couldn't say.

What are the average densities for each of the 4 lanes, and the number of raw and PF reads for each lane?

Our NextSeq generally maintains >90% PF with cluster densities up to 240K/mm^2, but the recommended spec by Illumina is 180-220K/mm^2. If you're outside of that range, I would recommend reducing your loading concentration proportionally based on how far over you are loading (i.e. if you're loading at what you believe to be 1.8pM, and getting densities of 270K/mm^2, try loading instead at 1.8 * (200/270) = 1.3 pM to target that 200K/mm^2 density.


One other possibility - Is your phasing/prephasing higher in lanes 3 and 4? The fact that both lanes seem to trail off in intensity (and presumably density) indicates a possible delivery issue. If the phasing and prephasing for those two lanes is higher than L1, that would support the idea of a reagent delivery/fluidics problem. If this is the case, only a service engineer call can guarantee restored functionality, but you could try doing a few full washes with warm MilliQ water to flush any residue/clogs from the lines.

[misterc]

NextSeq fluidics lane pairs are odd vs. even (i.e. lanes 1 & 3 get SBS pumping while lanes 2 & 4 are imaged).

[JoeKutch]

NextSeq fluidics lane pairs are odd vs. even (i.e. lanes 1 & 3 get SBS pumping while lanes 2 & 4 are imaged).

That's true, yes. But are lanes 1/3 and 2/4 treated as a single continuous lane with SBS flowing through 1 and then into 3, or does the line split and feed both of them? I always thought it was the latter, which doesn't preclude the individual lanes clogging. I could definitely be wrong about that though, I've only peeked inside our NextSeq during service calls, I haven't dug in it myself.

[misterc]

Good point. Yes, it's the latter (separate syringes pull the fluid through each of the two lanes simultaneously). It's entirely possible thiNGS has a fluidics issue on lane 1 there, but it seems like the fluidics on NextSeq are much more robust against clogs than it's HiSeq predecessor.