I still think the Nanopore people just have a tough product to fit into the market. This technology had much higher error rates than traditional sequencing (~15%) and the use cases where I want to analyze something out in the field so badly are not numerous. Collecting samples and sending them to your centralized sequencing facility is going to remain the most cost vs quality effective option. Additionally, sequencing is a huge overkill for a diagnostic test, especially when with the cost and complexity of analysis being so high.
Anyone have an idea of a market or research project where this thing is a first-choice tool?
Edit:
I stand corrected, seems some people are trying to use it in front-line clinical tests with post-analysis accuracy roughly on par with Illumina machines.
In my group (synthetic biology), we are using Nanopore MinION to do (a) Hybrid assembly and (b) Sequencing of bar codes/transcription units in iterative synthetic biology design.
For use case of hybrid assembly, we find that when you need not only an accurate but also contiguous denovo assembly. Hybrid assembly works really great because you're combining the best of both worlds, accurate short illumina sequencing and to span the gaps or ambiguous paths across contigs; use the Nanopore longer reads. This is important for use cases such as genomic surveillance of novel changes (e.g., synthetic engineering or horizontal or foreign gene transfer) - as you can't rely on a reference genome but need to detect a small change in the genome in an denovo fashion.
For use case of iterative synthetic biology design, we find that Illumina short reads don't cut it when you're doing a high throughput screen sequencing of a pool of say synthetic genes where they have similar barcodes, promoters or gene sequences (you're screening for which family of enzyme would work best, so you end up with sometime very homologous sequences). Because Illumina's 100-250bp doesn't span the entirety of a sequence and having a pool of very similar sequence makes the demultiplexing very tricky. We used Nanopore and the results were mixed given the high error rate of MinION.
Just sharing b/c it's been interesting to hear other people's application of Nanopore MinION!
Accuracy as you have noticed has been getting better, and keeps improving with their newer iterations, which include e.g. dual read heads in the nanopore.
> Anyone have an idea of a market or research project where this thing is a first-choice tool?
Plants. They often have huge genomes with sections that are hard to sequence with traditional methods with shorter read lengths.
I think that’s a use-case for Nanopore in general, but they explicitly call out that the app here is only useable against small genomes (Virus/Bacteria). I actually really prefer the movement away from optical reads and using smarter materials, some of the PacBio stuff is still optical but is focused on single cell work mediated by nano-materials as well which is interesting.
The part I was really interested in was why portability matters, instead of just having a bench-top Nanopore. I think the idea is that it will be more hobbyist friendly, would be fun to sequence random stuff while hiking. My concern is that the market for these devices isn’t actually that large, as any place doing meaningful analytics is probably going to want a bench top unit.
Looking at the prices, now, though I think $1,000 for the minION is about what I’d be willing to pay for that vs a $50,000 benchtop unit. The fact that it does RNA too is huge, as that has a lot of benefits to doing it as close to site of harvest.
I would love to have a little device that could reliably identify any plant. I'd teach myself to identify every plant in my area. I once counted fifty different wild species just in the ditch behind my house.
It's possible to learn from guidebooks, but I often find the descriptions a bit hard to interpret.
Plasmid sequencing is what I’m using it for. Nanopore is the only technology (other than Pacbio) that you can solve the indexing problem with, which MASSIVELY lowers the cost per sample. Approximately $14 down to $2, which a LOT less manual work (or a lot less robot work)
With plasmids, you can also train custom models to get better alignment results.
It’s insane that nobody else I know of is doing this, since it actually works quite well!
Are you sequencing the whole plasmid or just your inserts? The reason I'd never use a nanopore for my plasmid work is because the reason I sequence is to confirm assembly and sequence, especially after cloninf protocols, which is going to be one of the more common (if not most common) use cases for plasmid sequencing. I need Sanger for the base pair level accuracy.
In my experience, Sanger is less accurate than Nanopore, especially since you’re getting 1x coverage per read, and in order to reliably get good coverage of sequence you need a primer every 600 base pairs.
Consensus reads on Nanopore aren’t that bad, and you can automate them way better than Sanger. I haven’t found any issue using them - IMO the “it isn’t accurate” stuff isn’t very accurate.
Also, base pair resolution is bad on Nanopores primarily at homopolyer stretches. Guess what you can’t synthesize in the first place?
(Edit: I’ve done Sanger sequencing on about 3,000 plasmids at once before for high throughput cloning, and that SUCKED.)
So, can this be used to track where people have been? Follow the plume of DNA from the skin cells they shed? Is it an immediately useful way to tell if someone has COVID-19? Could it be used to covertly build a database with a unique DNA ID for all people? How fast is it, how much does it cost, are there consumables that have to be replaced? Some answers here:
Yea, you can use it for COVID19 diagnostic tests. I did that back in April-ish for a project before they redirected the Nanopore reagents to commercial COVID testing.
Yes, it could be used to build a database, so long as you have a good amount of DNA from the person. Honestly, governments are already doing this with Illumina runs, so no diff there really.
It immediately generates runs, and tapers off at about 12-24 hours, and using a full flow cell takes like 48. How long you run it for depends on depth of reads.
Flongle flow cells cost $99 and approximately $200 at end of day and minion flow cells cost about $1000, with minion being about 10x as good as flongle. The primary cost is definitely consumables, the capex for these devices is very low.
It's sad that our privacy and hence well-being is so intimately threatened by technological progress. It says something about the powers that be. In the ideal setup, we'd only be thrilled by new technology, while legislature and its enforcement would regulate things that might harm us, like tracking.
I didn't realize that they were this far with Nanopore technology. Really awesome stuff.
The weird thing in this article is that they say that "Most of the studying of DNA: aligning, analyzing, is done on large server clusters or high-end laptops." but that an iPhone app can replace that. In my experience, aligning is not computationally intensive at all.
The tricorder can typically do whatever it needs to, but it's normally portrayed as a super-powered radiation (meaning EM, but also Trek's made-up radiations) scanner.
That scanning seems to work down to the molecular level, because IIRC it's often used to check for "DNA degradation/mutation/etc", which is a common enough plot point. That implies the ability to take DNA sequences.
But this is just another case of a company trying to ride the Trek-tech bandwagon. A real "tricorder", or anything close, would be a massive leap forward in technology, even if it came in the bulky TOS-style boxes.
For those who didn't know, the mentioned "DNA sequencer" device is the smaller one on this page, the bigger already has a display and its biggest dimension matches iPhone X (but it's thicker and wider) so also relatively pocket-size:
They’re making the Smidgion which can directly connect to smartphones. The minion needs a USB connection tho, which is the $1000 one.
The raw data is actually pretty intense to decode - it took my 4 core i7 processor (2016) about 30 hours to basecall like 3gb of data from a flongle, which normally on minion flow cells you get like 30gb. This can be greatly increased with GPUs, however.
Nanopore is very locked down, but data you can share freely.
Anyone have an idea of a market or research project where this thing is a first-choice tool?
Edit: I stand corrected, seems some people are trying to use it in front-line clinical tests with post-analysis accuracy roughly on par with Illumina machines.
[0] https://www.nature.com/articles/s41587-019-0156-5
[1] https://www.nature.com/articles/s41591-020-1105-z