Core-facility NovaSeq users

[GenoMax]

I am interested in hearing from core-facilities that have NovaSeq's (or from other users where you don't always have control over library characteristics).

1. Do you pool libraries from multiple customers/sources? Do you insist on making the libraries or do you accept customer made libraries?
2. What QC measures do you use to ensure that you have a reasonably balanced starting pool/know what is actually in it (e.g. possible index submission errors from customers).
3. Do you run a MiSeq/iSeq diagnostic run before actual NovaSeq run?

Thanks in advance!

[nucacidhunter]

In my experience most places:

1- pool multiple customers libraries only for libraries made in the facility
2- accept customer made libraries for running in a lane or flow cell in customer risk for index or other issues. QC includes BA run and qPCR for quantification.
3- diagnostic run normally for non-standard libraries such as ATAC-Seq, Hi-C

[GenoMax]

Thanks @nucacidhunter. I suspected as much.

Massive capacity of S4 FC's for NovaSeq is posing a new challenge for cores. In order to get the maximum benefit of cost savings one needs to massively multiplex and we can't always control who makes the libraries and the quality (critical). While NovaSeq XP workflow is an option it adds a not insignificant cost to each NovaSeq run.

[pmiguel]

(1) Yes, we accept libraries from customers and pool them with other libraries. But the libraries have to be unique dual indexed and we do some QC to make sure they would not be a major source of index hops.

(2) MiSeq titration run. We can tell from that if the customer was on the wrong index pair.

(3) Yes. Not a full run, just enough to allow titration. Alas the MiSeq run is not as good a predictive tool for relative cluster densities for the NovaSeq as it was for the HiSeq. Our Illumina FAS mentioned there was some hope that the iSeq might be better as it also has a patterned flowcell. But we don't have an iSeq, so I guess that is a moot point for us...

Also, we heat denature libraries and run them on a pico RNA chip for QC, rather than running them undenatured on a DNA high sensitivity chip. This catches libraries with large amounts of primers/adapter dimers even if they are annealed to full length library molecules. In most cases Ampure fixes these. But we have had one strange case recently that always produced a trail, or even a peak of lower molecular weight entities even with extra Ampure. The Ampure seemed to make no difference.

Okay, this is a tangent, but possibly of interest:
These libraries were constructed by a customer who cut them out of a gel after amplification. They were using a UV transluminator and "Gel Red" stain for this. We loaned them a Dark Reader hand-held lamp to use for this purpose and the issue disappeared.

We never use a UV transluminator. I always advise those starting a new lab not to buy one and to buy a Dark Reader instead. I was shocked to learn that a large percentage of them had never even heard of a Dark Reader!

But to tell you the truth I'm still not sure what about the UV exposure caused the issue described above. Almost like it made the DNA "brittle", or caused cross-strand linkage formation. The lore is that UV causes "pyrimidine dimers". Which I would not expect to produce the issue that we saw.

--
Phillip

[sbarberan]

Hi Phillip,

Really interesting points about purification.

You mention that using the RNA chip allows you to see primers/dimers even if they are annealed to full length library molecules; and that if this is the case you can purify them with AMPure.

My question is, if they are annealed to full length library molecules how come AMPure is able to remove them? AMPure will not denature the libraries so if they are annealed they will be co-purified, right? am I missing something?

Cheers,
sergio

[GenoMax]

(2) MiSeq titration run. We can tell from that if the customer was on the wrong index pair.

(3) Yes. Not a full run, just enough to allow titration. Alas the MiSeq run is not as good a predictive tool for relative cluster densities for the NovaSeq as it was for the HiSeq. Our Illumina FAS mentioned there was some hope that the iSeq might be better as it also has a patterned flowcell. But we don't have an iSeq, so I guess that is a moot point for us...

--
Phillip

Hi Phillip,

Thank you for the useful information.

Can you clarify what do you mean by just enough to allow titration? You don't let the MiSeq run finish (I assume you are using 50x8x8 run)? You run the "pool of pools" that is going to be loaded on NovaSeq and not individual customer pools, correct?

AFAIK iSeq reagent costs are very similar to MiSeq so other than the possibility of having a platform with patterned wells it does not appear to have any major advantage. Illumina is also not discounting iSeq reagents for the foreseeable future.

[pmiguel]

Hi Phillip,

Really interesting points about purification.

You mention that using the RNA chip allows you to see primers/dimers even if they are annealed to full length library molecules; and that if this is the case you can purify them with AMPure.

My question is, if they are annealed to full length library molecules how come AMPure is able to remove them? AMPure will not denature the libraries so if they are annealed they will be co-purified, right? am I missing something?

Cheers,
sergio

Hi Sergio,
I agree. I would not expect it to work, but it does. We've tried a denature+snap cool prior to ampure and it performs the same as a non-pre-denatured ampure.
I have no explanation at the juncture, just giving the results we see.

--
Phillip

[pmiguel]

Hi Phillip,

Thank you for the useful information.

Can you clarify what do you mean by just enough to allow titration? You don't let the MiSeq run finish (I assume you are using 50x8x8 run)? You run the "pool of pools" that is going to be loaded on NovaSeq and not individual customer pools, correct?

AFAIK iSeq reagent costs are very similar to MiSeq so other than the possibility of having a platform with patterned wells it does not appear to have any major advantage. Illumina is also not discounting iSeq reagents for the foreseeable future.

We just create a pool using 1 ul of each library to be tested. But when we analyze the results we are looking at relative clusters/ul from libraries of the same type. That info is used to make a pool and that pool is actually quantitated by qPCR.

We sell fractional MiSeq runs -- down to 1% of a run. So many of our MiSeq runs have samples from many labs on it. So we can use 5-10% of the run for titration of samples. Most often these would be MiSeq 500 or 300 cycle runs.

---Begin tangential rant:
I like having a little 2x250 information for an RNAseq library -- you can catch cases where there is high ribosomal contamination or something else going wrong. And I almost invariably get asked by people processing the data afterward "what is the insert size of the library?" Which drives me nuts because (1) they don't want to know the insert size of the library, they want to know the insert sizes of the library molecules that clustered and (2) why ask me? They have paired-end data, all they need to do is map 100K reads to 28S rRNA for their species and do the calculation themselves. Probably more importantly, why do RNAseq packages ask for the average insert size of a library? Refer to points 1 and 2 above programmers!
Anyway, since we can pair merge the read pairs, adapter trim them and run them through fastqc. Then you can actually read the average insert size off the fastqc "read length" chart. As long as your inserts aren't too big. But generally RNAseq library inserts are short enough for this to work.
---End tangential rant.

Anyway, MiSeq titration still sucks compared to using one NovaSeq run to titrate another. But that isn't often an option. And the MiSeq method seems to work as well as qPCR does (not great) and isn't as laborious. We just have a little ABI Step-One -- 48 wells total. So it isn't really ideal for processing large numbers of samples. The standards eat up enough of the plate that we can only fit 11 samples per run if we do them in triplicate.

The main limitation of the iSeq is that it doesn't, as far as I know, have a 500 cycle kit. Most amplicon experiments people do use the 500 cycle kit. If it did have a 500 cycle kit I would seriously consider trying to obtain an instrument. Well, if is shown to actually provide better titration info than the MiSeq. Which it might not.

--
Phillip