[pmiguel]

We have some TruSeq libraries that over clustered recently and it looks like we need to run them at a much lower concentration. Our FAS suggested to reduce concentration by at least 30%.
We've requantified again and it seems our Qubit and Bioanalyser traces were correct. We even double checked by qPCR against a library we already sequenced and results were similar. We're not sure why these libraries are over clustering.

What type of qPCR? If SYBR green you can still run into adapter dimer issues. See this for example of some libraries that show absolutely no adapter dimer on a high sensitivity DNA chip, but some when strand denatured and run on an RNA pico chip. The library portrayed in the link above was fine -- the adapter dimer amounts were low. But we have seen other libraries where the adapter dimers (or maybe they were primer dimers) were a substantial molar contributor to the library. As a result we overclustered them by quite a bit initially.

Since SYBR green qPCR relies on an average construct size to determine concentration, having even 10% (by mass) of your library as adapter dimer can really through your clustering off.

Oh, also we noticed that ethidium bromide seemed to wreck picogreen fluorimetry entirely and appeared to throw off qPCR as well. We frequently do a Pippin prep size selection on libraries, so this is an issue for us.

Finally, you are aware that something about the TruSeq library method causes them to out "perform" other libraries as far as clusters/pmol of library added, since you mention this above. Possibly the extra magic comes some modification to the enrichment PCR primers?

--
Phillip

[epistatic]

My lab has recently run into this over clustering issue with 10pM.

At first I thought we didn't quantify our libraries correctly. I ran Kapa and Agilent qPCR kits on a set of samples and feel our quantification was correct. Fresh 2 nM dilutions made using qPCR results and still gave too high clusters.

What we think is happening is that through HiSeq software upgrades, more clusters are being called but we are not being told to adjust our input concentration to still hit the target number of clusters. We can all discuss the initial quantitation methods but assume if the sample pM is correct there can be another factor (ie cluster calling). I think the cluster densities are becoming a moving target. I discussed with Illumina that pulling more data out of the flow cell is only helpful if it is quality data, not if we still should always hit less than ~800K/mm2.

Our interpretation could be totally off, I would be happy to hear others experience. There is also the thought that some over PCR'd RNA samples have a population of ssDNA that don't qPCR quantify correctly (comparing to a fully dsDNA ladder) but still seed clusters. It could be a combination of factors, but I think the suggested pM load should be reduced.

GAIIx = 4-7 pM
HiSeq 2000 v2 = 6-8 pM
HiSeq 2000 v3 = intially ran 11-12 pM, were told to try 10-14 pM. That seemed ok until recent runs with new software.
HiSeq 2000 v3 with recent software upgrades = 12, 10, 9 pM too high with 9 pM ok for tight band pippin'd DNA but not RNA. Going back to try 8 pM.

[pmiguel]

ssDNA may show in a DNA Bioanalyser trace but run faster.

It stands to reason and experience with other electrophoresis platforms. But this did not seem to be the case when we tried it. Almost looks like bioanalyzer chips are "tuned" to yield approximately the same length whether running single stranded or double stranded polynucleotides--at least at the smaller sizes. But it looks like somewhere above 1000 nucleotides the single stranded molecules run much, much more slowly (larger) than double stranded molecules.

--
Phillip

[pmiguel]

But why would a qPCR show a higher concentration than what Bioanalyzer does? I mean, what DNA would Bioanalyzer not take into account?

Realtime PCR?
Short (adapter dimer or primer dimer) single stranded molecules that have annealed to a longer (library) template molecule. Even SYBR green qPCR may be confounded by this effect. See this for a case we ran into. Note that the bioanalyzer trace in the middle of the post is strand denatured library run on an RNA pico chip. Hence it was possible to visualize the, otherwise hidden, dimers.

--
Phillip

[Susanne]

We tested ssDNA (primer size, <100 nt) on the Bioanalyzer DNA assays, and it does run slower than dsDNA of the same size (roughly -25-30%).
Also, the quantification for the ssDNA is off (may be around 50%).
So, for sure you are able to see ssDNA on the Bioanalyzer DNA assays.

[pmiguel]

We tested ssDNA (primer size, <100 nt) on the Bioanalyzer DNA assays, and it does run slower than dsDNA of the same size (roughly -25-30%).
Also, the quantification for the ssDNA is off (may be around 50%).
So, for sure you are able to see ssDNA on the Bioanalyzer DNA assays.

Thanks for your post Susanne! You are the first person here to confirm the observation I made quite some time ago! Well, ECO did chastise me for claiming that this was an unexpected result. So that was acknowledgment, of a sort.

But I have been waiting for someone to either confirm or refute this result.
--
Phillip

[pmiguel]

My lab has recently run into this over clustering issue with 10pM.

At first I thought we didn't quantify our libraries correctly. I ran Kapa and Agilent qPCR kits on a set of samples and feel our quantification was correct. Fresh 2 nM dilutions made using qPCR results and still gave too high clusters.

What we think is happening is that through HiSeq software upgrades, more clusters are being called but we are not being told to adjust our input concentration to still hit the target number of clusters. We can all discuss the initial quantitation methods but assume if the sample pM is correct there can be another factor (ie cluster calling). I think the cluster densities are becoming a moving target. I discussed with Illumina that pulling more data out of the flow cell is only helpful if it is quality data, not if we still should always hit less than ~800K/mm2.

Our interpretation could be totally off, I would be happy to hear others experience. There is also the thought that some over PCR'd RNA samples have a population of ssDNA that don't qPCR quantify correctly (comparing to a fully dsDNA ladder) but still seed clusters. It could be a combination of factors, but I think the suggested pM load should be reduced.

GAIIx = 4-7 pM
HiSeq 2000 v2 = 6-8 pM
HiSeq 2000 v3 = intially ran 11-12 pM, were told to try 10-14 pM. That seemed ok until recent runs with new software.
HiSeq 2000 v3 with recent software upgrades = 12, 10, 9 pM too high with 9 pM ok for tight band pippin'd DNA but not RNA. Going back to try 8 pM.

Yeah, I think something is up. Could it be a change in the cBot sofware? We clustered at 14 pM -- which was giving us 600-900 Kclusters/mm^2 with previous runs. The density of our run started yesterday was right around 1000! This was from 2 nM lane pools that were qPCR (KAPA) checked for concentration with phiX v3 as a standard then adjusted prior to further dilution.

Wait, these may also have been v2 TruSeq libraries. I wonder if Illumina put extra "magic" in their PPC for v2? v1 PPC was already producing amplicons that out clustered generic oligo primed libraries. Back to 10-12 pM, I guess. Oh well, no big deal Illumina. A new PE cluster kit is only $5000...

--
Phillip