[niceday]

We do get qPCR results higher than Qubit and bioanalyser. I spoke to some-one at Kapa and they did say others had observed this. We use the qPCR as standard and we get reproducible results.

[HESmith]

I am wondering is any one observed that the qPCR reading is higher than the Qubit? I mean this is impossible right as the qPCR only measure the molecule which ligate with adapter.

The Qubit is specific for dsDNA. If you sample contains ssDNA with adapter ends, it will be detectable by qPCR but not Qubit. (Note: it will also form clusters, which makes qPCR more accurate for quantification.)

[pmiguel]

The Qubit is specific for dsDNA. If you sample contains ssDNA with adapter ends, it will be detectable by qPCR but not Qubit. (Note: it will also form clusters, which makes qPCR more accurate for quantification.)

In addition to ssDNA amplicons in your library, adapter dimers (ssDNA or dsDNA) could also be a source of higher molar concentration readings from qPCR versus fluorimetry.

Adapter dimers (zero or very short insert amplicons) will form clusters but will be detected at different sensitivities by different QC assay methods. Probe-based qPCR would likely detect them accurately. SYBR-green based qPCR might underestimate their molar concentration.

As detailed up-thread, Agilent 2100 Bioanalyzer results will detect non-amplicons (one or zero adapter ends) the same as amplicons -- other than the (probably minor) shift in length. Other confounding aspects include:

(1) ssDNA runs slower than one might expect on a High Sensitivity DNA Bioanalyzer chip. (Probably also on other DNA Bioanalyzer chips). Run a lane of nano RNA ladder on your DNA chip to get an idea. Anyway, this can be confusing under some circumstances.

(2) dsDNA runs faster than one might expect on RNA (single-stranded) Bioanalyzer chips. You can denature your sample (95 oC, 2 minutes followed by "snap chill" on ice) prior to loading, but the sometimes crazy results can be hard to interpret. Also, you may end up wondering whether your sample renatured, despite your efforts to prevent this from happening.

As far as fluorimetry goes, there are also single-stranded fluors, like ribo green, that can be used. But this can be a mine-field as well. The main, perhaps unexpected, issue is that ribo green fluoresces somewhat more brightly in the presence of double stranded DNA than single stranded. No where near the difference one sees with pico green, but still ~2x. We currently are trying to use ribo green, but using double-stranded DNA as our standard (phage lambda). We are denaturing our standard and our samples using the method mentioned above.

My assessment: all the library QC assays have significant issues. Beware.

--
Phillip

[kulukulas]

(1) ssDNA runs slower than one might expect on a High Sensitivity DNA Bioanalyzer chip. (Probably also on other DNA Bioanalyzer chips). Run a lane of nano RNA ladder on your DNA chip to get an idea. Anyway, this can be confusing under some circumstances.

do you expect to detect ssDNA in DNA high sensitivity assay at all? isn't it a dsDNA binding dye used there?

[pmiguel]

do you expect to detect ssDNA in DNA high sensitivity assay at all? isn't it a dsDNA binding dye used there?

Regardless of what my expectations were, we have run the ssRNA ladder on a DNA high sensitivity chip. We did detect it.

See:

http://seqanswers.com/forums/showthread.php?t=12852

for details.

I should add that when we asked Agilent (prior to running this chip) what results we should expect when running single stranded molecules on a double stranded chip, we were told that the chip was not appropriate for that purpose. (Thanks for nothing, Agilent...)

--
Phillip

[niceday]

to ktabbada

If you are seeing qPCR results that are 10% of your bioanalyser or Qubit readings and you successfully clustered using the qPCR concs then you should have your bioinformaticians scan the data to see if you have low complexity library.

Also are you making PCR or nonPCR libraries?

With PCR after adapter ligation properly ligated material should drown out incorrectly ligated library and your bioanalyser/Qubit should be close to your qPCR conc. If you see a big difference between bioanalyser/Qubit and qPCR you might have low ligation efficiency or poor PCR amplification.

On another note I have seen qPCR data higher than both ubit and bioanalyser. I always use qPCR and cluster accordingly.

[HGENETIC]

We've always used picogreen and found it matches well with the qbit results that our sequence service provider uses to make library dilutions before running. However, after some curious results we are going to try the Agillent NGS QPCR system as this is supposed to be more accurate, just have to wait to see how the results tie up.

[sehrrot]

I agree with Lee

we used KAPA, this is the best option for cluster density. We've succefully used it for V2 and V3 SBS kit.

[Genquest]

Library qualification by qPCR

Hello everyone,
I am pretty new to this forum and stumbled on this post as I am facing problem with qPCR quantification method
1.Has anyone encountered with, a higher qPCR concentration than the estimated concentration by pico green??
2. Also if assumed that qPCR is more accurate(w.r.t pico green) as it would give conc. of ssDNA ligated with adapters,then the cluster generation should be accurate with qPCR conc. but it doesn't happen so...the flow cell (V3,Hiseq 2000) remains under clustered.Can any one think of possible reasons??
3. This trend seems to change with every library.Is absolute quantification the right approach for Sybr green qPCR method?Also I am using the PhiX control provided by Illumina as my standard.
4.Do people take the "exact concentration of libraries" as determined by qPCR for cluster generation or they compare it with pico green to see the relative trend of concentration?

Would appreciate if any of the queries are resolved!!
Thanks in advance

[HGENETIC]

Hello everyone,
I am pretty new to this forum and stumbled on this post as I am facing problem with qPCR quantification method
1.Has anyone encountered with, a higher qPCR concentration than the estimated concentration by pico green??
2. Also if assumed that qPCR is more accurate(w.r.t pico green) as it would give conc. of ssDNA ligated with adapters,then the cluster generation should be accurate with qPCR conc. but it doesn't happen so...the flow cell (V3,Hiseq 2000) remains under clustered.Can any one think of possible reasons??
3. This trend seems to change with every library.Is absolute quantification the right approach for Sybr green qPCR method?Also I am using the PhiX control provided by Illumina as my standard.
4.Do people take the "exact concentration of libraries" as determined by qPCR for cluster generation or they compare it with pico green to see the relative trend of concentration?

Would appreciate if any of the queries are resolved!!
Thanks in advance

Hi Genquest,

we had a similar situation to yours in that our qPCR results gave higher DNA concentration than our picogreen (see post above for qPCR kit used). However, when ouor sequence provider ran a qbit the results were more in line with our picogreen and so we decided to use that concentration for cluster generation. When the samples were run we had a very high cluster density and it looked like the qPCR results may of been the most accurate, luckily we still got good data but I can't explain why the two methods gave the results they did and we haven't run anymore samples so I can't comment on whether this was a chance result. My advice would be use one method (preferably qPCR), optimise it and stick with it. Cheers H

[Genquest]

Hi HGENETIC,

Thank you for your reply!!I feel relieved to know that others also have experienced such problem).
What happened in my case was just the reverse....We took the qPCR concentration for cluster generation (for V3 flow cell on hiseq2000) assuming that the qPCR would give accurate results and landed with under clustering.

I guess we just cannot take the qPCR results straightforward...have to see the trend.Need some advise on optimization!!
Can anyone help please??

Thanks in advance!!

[TonyBrooks]

I just saw exactly the same thing. I've run hundreds of samples on our GAIIx's but we thought we'd switch to qPCR after upgrading to the HiSeq.
I used Qubit values to normalise to what I thought was 10nM. When I ran those on qPCR I was getting anything from 5nM to 30nM.
I used both the standard in the KAPA kit and a homebrew (from Illumina PhIX). Both standard curves had R2>0.999. The efficiency of the PhiX curve was 104%, but the KAPA had much more variable efficiency (80-120%).
The kit standard gave me ~30% lower concentration values than the PhiX.
I went with the PhiX as I'd run that on the HiSeq during training so I knew the expected cluster density was good at 17pM, I didn't want to over cluster and there were strange efficiency values for the Kapa curve.
However, when I sequenced, the density was much lower than expected (~500k/mm2). The KAPA standards would have given more accurate quantification.
Finally, to compound matters, we had one sample in a multiplex pool give twice the number of reads than we would expect if everything was equimolar. So it looks like it was under-quantified by qPCR.
So although qPCR seems to be better than Qubit, it's not without its problems.
I would be interested if anyone has some ideas on how to optimise to increase accuracy too.

[sehrrot]

hi all

1. Kapa kit is the best option (at least to me and my colleagues). (to Tony) please check your Kapa kit's standard, then you will get nearly 100% curve (as I always did).
2. optimising the optimal cluster is a sort of relative thing. Once you have training run, you will need to run PhiX to find a optimal range of the cluster density (recommend to try few pM ranges): so I cannot say the absolute value of this.
3. what I concern for Genquest's question is,,, what was you library type? whole genome? exome? RNAseq? My feeling is that,, ,whenever I did RNAseq, its qPCR always gave a variable range of actual molarity. For example, I measured an RNAseq sample as 10nM but it came as 90nM. I am pretty sure this variable would be attributed to variable length of library, which is also involved in calculating the molarity. So, for RNAseq library, I really recommend to measure the "exact size" of the library
3-1. For RNAseq library, you should be careful about contamination of probes or primer dimers. Even if you see very tiny portion of this in bioanalyzer peak, those will cause serious problem in sequencing, I mean "decreasing library complexity". Of course, I pretty assume that those tiny fragments will affect on the molarity of a library.
3-2. Regarding whole genome & exome library, you wouldn't be worried about qPCR things. If you still get bizarre result in qPCR, you will need to check either your pico-green or qPCR method.
4. I haven't used Illumina qPCR protocol for long time since I got Kapa kit. It's bit sad that Illumina qPCR protocol has the weird description for serial dilution of a library (it might be corrected now though..)

[sehrrot]

One more quick post,

A bit different behaviour between GAII and Hiseq in terms of finding the cluster density and actual showing on the cluster, me and my colleague think..

[pmiguel]

Looks like quite a few discussions of this issue throughout the forum today. Seems to me that we want to distinguish between "major" and "minor" failures to correctly titrate a library. The cut-off between the two would be arbitrary. But I would say anything less than a 2x difference in what you thought the cluster density would be and what it actually would be reasonable.

If you are consistently getting within 2-fold of where you thought you were, you are in a completely different situation than if you are (for example) 9x wrong. Less than 2x suggest that your assays need a little tweaking. Maybe check to see if you just have a systematic bias that can be corrected using a static factor. Or if things are random, then looking for ways to reduce the noise in your system are called for.

Whereas if you are >10x off then you are looking more towards catastrophic issues.

Even though I frequently rail against people not distinguishing between an assay being 30% off or 300% off, I was recent bitten by failing to do exactly this. I had a flow cell badly over cluster. In the throws of grief and horror, I remembered that I had failed to correct for the longer length of my libraries with respect to the standard. This came to as much as a 1.5x correction factor.

We re-clustered using the new factor. Still over-clustered! Turns out there were a host of other issues with our estimations, the greatest of which was that our phiX standard concentration was wrong!

Anyway I have seen a "when it rains, it pours" phenomenon occur on more than one occasion. That is, something that used to work, no longer does. But when one looks into it, there was not one point of failure but a whole host of them. I think this derives from an attempt to keep our methods agile and costs low -- most standard QC is discarded. But also, it deals with the robustness of the method itself. If a set of methodology is fairly robust, no single minor issue is likely to produce systemic failures. So by the time you are troubleshooting, you are not tweaking or correcting a minor issue, you are dealing with a multitude of issues.

--
Phillip