Pooling for Oxford Nanopore Multiplex Sequencing

[hathormaat]

Good Morning All,

I am trying to sequence a few hard-to-ID Enterobacteria using the Oxford Nanopore Rapid Barcoding Kit (SQK-RBK004) on the MinION. These samples have already been sequenced on an Illumina Miseq.Through trial and error I've found that ONT's promise of 12x multiplexing is...optimistic at best. To maximize reads I've cut multiplexing down to 2x and seem to get decent reads with good quality (for ONT), although even still I can't seem to get good read-equity between samples. One might have 1.2Gbp called while the other has 1/4th that number.

My workflow is slightly modified from the base protocol as follows:

-Pre-Library-Prep Bead concentratio using Axygen Ampure beads. (1:1 Ratio with 30 minute incubation + 30 min 37C Elution). Our samples are not nearly as concentrated as needed for the protocol so this is a must for library prep to even have a hope of working.

-2x initial rxn volume (15uL gDNA input, 5uL Fragmentation Barcode)
(10uL was too small to ensure volume sufficient for Qubit / Tape Station QC)

-Post-Library-Prep Bead Concentration to pool samples to the 10uL volume required for loading the final library.


For pooling, I've followed the approach that we use for our Illumina runs.
-Qubit fluorometer to get concentration
-Agilent tape station to give fragment size
-Dilution excel sheet to calculate how much volume is needed to ensure an equimolar amount between samples.


So far it seems as though my attempts at normalizing my PAL are a crap-shoot with no real correlation between my efforts and read equity. This is a PCR-Free protocol, so perhaps the lack of a single amplified fragment size is throwing my assumptions of "Average Fragment length" out the window.


Anyone have any suggestions? Protocols? Suggestions? Experienced something similar?

Thanks!

[gringer]

Assuming you're doing this on a MinION flow cell, you can do an initial run on a Flongle flow cell to work out barcode distribution, and use that to balance the pool for MinION sequencing. We are doing this now with all our cDNA runs.

Especially when extracting from diverse sample types, actual results can vary wildly from what is detected via Qubit / TapeStation.

[hathormaat]

Is there any literature you all used to base your flongle experiment? I'm interested to see if this could work for us. Because we're using the Rapid Barcoding Kit on gDNA samples with relatively low mass I'm worried about losing DNA that would be better used on the actual run.

Could you elaborate on Qubit / TapeStation differences in detection among diverse sample types? While our samples come from diverse genera of bacteria, they are all prepped, cultured, and extracted following the same protocol such that all isolates are gDNA.

thanks!

[hathormaat]

Assuming you're doing this on a MinION flow cell, you can do an initial run on a Flongle flow cell to work out barcode distribution, and use that to balance the pool for MinION sequencing. We are doing this now with all our cDNA runs.

Especially when extracting from diverse sample types, actual results can vary wildly from what is detected via Qubit / TapeStation.

Is there any literature you all used to base your flongle experiment? I'm interested to see if this could work for us. Because we're using the Rapid Barcoding Kit on gDNA samples with relatively low mass I'm worried about losing DNA that would be better used on the actual run.

Could you elaborate on Qubit / TapeStation differences in detection among diverse sample types? While our samples come from diverse genera of bacteria, they are all prepped, cultured, and extracted following the same protocol such that all isolates are gDNA.

thanks!

[gringer]

Literature? I don't understand what you mean by that.

I've noticed across many different runs that sequencing run performance on nanopore for the same prepared and adapted library has near-identical barcode proportions regardless of run time or platform. This is expected, given that it's essentially a random sampling device. Previously, we were doing a short 15 minute run on MinION flow cells to work out actual running barcode proportion, and using that to rebalance pools and resume sequencing with a better mix. Now, we use Flongle flow cells for the same purpose. It's a little bit more expensive in terms of run cost, but there's less sequencing wasted on poor samples, so I think it works out cheaper in terms of cost per read. It'll work regardless of the sample preparation: ligation, RNA, rapid Barcoding, cDNA,... whatever.

Nanopore sequencing is heavily affected by short reads: they're small (so move faster), and have greater numbers for the same mass (so take up more adapter during sample prep). Unfortunately, they don't show up so well on a TapeStation or gel because they have low mass, so you usually don't know how many there are until you're 10 minutes into a sequencing run.

Qubit quantification (or Quantus) is a reasonable representation of mass, but doesn't help much with molarity (which is what matters more for Nanopore sequencing). It can also be affected by contamination, un-adapted template, and RNA (if RNA concentration is sufficiently high, at least ~100x DNA concentration). It's fine for a rough guess for balancing barcodes, but the best and most repeatable statistic to use is actual read counts and proportions from a sequencing run.