View Single Post
Old 03-08-2016, 02:03 PM   #28
Location: Bay area

Join Date: Jun 2012
Posts: 77

Originally Posted by DNA_Dan View Post
Assuming the libraries range across the board, Truseq, Nextera, Kapa Biosystems, Nugen, etc., come from a huge variety of DNA/RNA sources, and have different fragment lengths/profiles - How tight can the ddPCR dial them in?
Within chemistries, pretty good. We can certainly dial in 1000K/mm2 when the libraries are produced with a standardized workflow.

To elaborate, with the Kapa Biosystems qPCR method the tightest we can normalize our pools of samples is 1-3 fold of each other. That is to say sometimes they are pretty even, other times they can vary by up to 3-fold. Can you achieve a better normalization in a pool of very different libraries with completely difference efficiencies using the ddPCR method? If so, how tight? 80% or better? 90% or better?
I think that within libraries from a given chemistry, you should be able to be within 20% of each other. The biggest reason this is the case is that you are measuring number of molecules directly, not inferring from a standard curve with a size correction. The other advantage is that you can get a better handle of concentration in the presence of primer/adaptor carryover, which you just cannot from qPCR. You can also see whether you likely have empty libraries or one with nice inserts.

Cost is a huge pill to swallow because 90K is a lot of rounds of normalization kits and technician time. However I have tried a reiterative process of normalizing, measuring with qPCR, then normalizing again, 2-3 times one after the other and what I have found is that the pipetting error in the dilutions and measurement error have a limit with how close you can "dial" samples with respect to each other. At some point the pools don't get any more "normalized", they actually start to get worse because of the handling error involved or the measurement itself.
Absolutely. The 90K price is steep. Particularly if it was only used for library quant. But that's really not the point of the instrument.

And absolutely. Any measurement is only as good as the technique used to create the sample to be analyzed. ddPCR is not a panacea of accuracy. I played with a bunch of methods for creating those dilution series, but at the end of the day well-calibrated pipettors and consistent operator pipetting of 1:100 dilutions were the key. And if there is a consistent bias in technique, that can be accommodated in the calculations.

So in essence what I am looking for is something that has the accuracy to push the flowcell to it's maximum density reproducibly every time and normalize the pools so evenly you are squeezing every bit of data possible for each sample on every run. We also do a lot of ratio pools (30% one customer, 70% another) that sort of thing. Being able to target this accurately down to 1% would allow us to put more customer samples on a run because we would have the confidence that we would hit our ratio targets more accurately.

Is the QX200 the instrument that can do this? Is this a pipe dream? Where do you feel the QX200 falls short of expectations? What are its limitations?

Ultimately if the method allows for a tighter multiplex/run scenario as I described above, the cost savings of not having to run another run on the sequencer would pay for itself. The main cost and driving factor is still the sequencing reagents. is a pipe-dream to believe that any analytical measurement with manual handling steps can have 1% variance. Error is compounded over the number of handling steps.

If you are a high throughput lab, and you can use a full plate more often than not, the cost per sample is very reasonable. The minimally optimal format cost-wise is in groups of 8. If you do a lot of one offs, the cost per sample becomes higher, mainly because of the peripheral consumable costs. On that scale the QuantStudio may be better.

From a workflow perspective, once you get the hang of it, it is trivial. And if all you have to do is get the right normalization is do a single dilution series and a single measurement (or two...I usually do 1e-6 and 1e-4), then that also has value.


austinso is offline   Reply With Quote