Thread by @bentemperton

Our paper on using @nanopore MinION to do long read viral metagenomics is out today in @thePeerJ: (peerj.com/articles/6800/).
Here's a quick thread on what drove this research: The reason for its inception boils down to this: I love viral metagenomics, but hate assembly.

Taking environmental DNA, smashing it into a billion pieces, then trying to reassemble the jigsaw is an enormous fudge to deal with the limits of current technology.

Imagine your perfect environmental genomics sampling experiment. You sample your organisms, you extract your DNA from each organism intact and then you sequence it all and get on with understanding the biology.

That's not to say the metagenomics work being done isn't awesome - we've got very good at jumping through the hoops - I just think in 20 years time we'll look back and laugh at our current methods.
imgflip.com/i/2zesp4

If you think it's bad for the cellular fraction, then you'll not be surprised to hear that for viruses, it's no picnic either.

In the last few years, we've come an awful long way to solving this riddle, in no small part due to the sterling efforts of @simroux_virus, @gregory_annc and others in the @Lab_Sullivan (e.g. nature.com/articles/natur…), but there are definite pitfalls in current practices.

Not least, some of the most abundant viruses on Earth (see nature.com/articles/natur…), infecting a marine host that is responsible for returning a vast quantity of global primary production back to atmospheric CO2 (see dx.doi.org/10.1146/annure…) are poorly represented in assemblies.

Their lack of representation seems to be due to the fact that they fall apart during assembly and are thus discarded in the trash of contig fragments deemed too small to process (see peerj.com/articles/3817/ and nature.com/articles/ncomm…).

These problems can be circumvented to some degree with single-viral and single cell genomic approaches (nature.com/articles/s4139…) and fosmid libraries (journals.plos.org/plosgenetics/a…) but

MinION long reads appeared to promise a way to potentially give the benefits of these alternative approaches, but with sufficiently high throughput to make the analysis cost-effective, giving us an insight into these important viruses.

The downside? Most viromes from 20L of seawater yield several orders of magnitude too little DNA to run efficiently on a MinION.

In steps @JoWarwickD as my first PhD student, with support from @NERCscience, @BIOS and @royalsociety to take up the challenge of how to amplify viral DNA enough for sequencing on the MinION, whilst retaining relative abundance information

So we joined the @nanopore MAP (ahh, remember those days?) and went to Porecamp to learn how to use it under the guidance of @pathogenomenick , @DrT1973, @mattloose and @Justin_OGrady. As a new PI, I lucked out with @JoWarwickD joining my team because she absolutely smashed it.

Following tests on mock communities supplied by the @Lab_Sullivan, she developed a protocol to capture long read viral fragments called 'VirION' using an adapted linker-amplified shotgun library approach (protocol here: protocols.io/view/virion-lo…)

In this time we experienced 2 chemistry changes, a pore chemistry change and the demise of 2D sequencing and learned pretty quick that speed of developments for MinION are awesome unless you're trying to develop a protocol.

Once we had the data, we needed to figure out how to deal with the error rate that was screwing with gene calling (see nature.com/articles/s4158…) and thus the identification of viruses.

Combining it with short reads, we developed a hybrid assembly pipeline to maximise the benefits of both types of data

The end result is that in taking this approach, we discovered two things: (1) the long read viral assemblies yield a whole bunch of stuff that is missed by short read assemblies. Stuff that happens to be really abundant in the marine environment

(2) You can use long reads to peer inside viral hypervariable regions to see whether their variability occurs at the nucleotide or protein level (it's the former in the viruses we looked at).

So, we've taken a big step towards improving viral metagenomics, but we're not quite at the future yet where we don't have to assemble.

We're working on that now and are developing protocols that allow sequencing of full length viral genomes (and their DNA modifications) across a range of DNA input concentrations. Watch for those in a follow-up paper.

Once those are finalised, we'll be able to look at how viral communities evolve with unprecedented resolution in not just marine systems but also in phage therapy, agriculture and beyond.

For more info, you can check out my @MicroSeminar video where I discuss this work: .