In case you don’t know, the PromethION is the bigger brother of the MinION, Oxford Nanopore’s revolutionary mobile single-molecule DNA sequencer. The PromethION is about to go into early access, where I believe the access fee is just $75k, most of which can subsequently be spent on reagents/flowcells in the ONT shop. The hidden costs are of course the bioinformatics, but that’s not what this post is about.
What I wanted to do is predict PromethION throughput using current MinION stats. So let’s do that.
I will be deliberately conservative with the MinION throughput stats. It’s a matter of record, I think, that a MinION 48hr run is easily capable of producing 100-200Mbase of 2D reads, most of which are ~8Kb in length. I’m going to use this figure as a basic MinION throughput, though many will argue that SQK-MAP-006 runs produce far more.
I want you to focus on three parameters: the rate at which DNA translocates through the pores, currently ~70b/s on MinION; the number of channels per flowcell (512 on MinION); and the number of flowcells per run (1 on MinION). So 1 flowcell with 512 channels running at 70b/s will produce 100-200Mbase of high quality 2D sequence data.
PromethION flowcells are 4 times the size of MinION flowcells, with 2048 channels. You can also run 48 flowcells in parallel. Therefore, if all PromethION ends up being is 48 large MinIONs running in parallel, we can predict 4 * 48 * 100-200Mbase == between 19.2 and 38.4Gbase per run.
However, let’s go back to that translocation speed. I believe the aim is for the PromethION to run at 400b/s. There is a huge caveat here in that we haven’t seen any data from ONT running at 400b/s; however, logically the faster the DNA moves through the pore, the more sequence data you will get. The potential is therefore for PromethION to produce 5 times the above – so between 96Gbase and 192Gbase of high quality 2D reads (note – many more template and complement reads will be produced). (2nd note – there is no guarantee the 2D protocol will be around long term, the aim is to produce 1D data at a much higher quality, which will also increase throughput)
These are obviously “back of an envelope” calculations, and almost certainly wrong in some respect, but hopefully this puts the PromethION in context with the other sequencers on the market.
Update 1: via Nick Loman, the F1000 data was SQK-MAP-005 and ran at 30b/s, not 70
Update 2: via Matt Loose, PromethION flowcells are 3000 channels not 2048
Update 3: via Nick Loman, PoreCamp (SQK-MAP-006) run produced 460Mbase 2D data
* bad maths is a product of lack of sleep
** the blog is called opiniomics for a reason