Sometimes interchangeable, but I, I am inclined to agree with that.

Yeah.

And, and machine learning is, is of course huge.

And I would, I would, probably say most fields of science, all fields of science, perhaps.

I don't think you can avoid it in 2025.

And of course, the AI algorithms in particular have seen such a wild improvement in the last decade or so that it has absolutely revolutionized some fields.

And so, yeah, I think there are, countless places to use machine learning and AI in microbial genomics.

There's acetylation, there's additional methylations that then they at least had models for last I checked.

So how did they catch your attention as opposed to some of the others that were available?

So yeah, I was working in Kat Holt's lab at the time.

I think this would have been after I finished my master's, but I was a research assistant in her lab.

And Kat Holt's lab works with a lot of Klebsiella, which are a species of bacteria that often has a lot of plasmids.

It's not uncommon for a Klebsiella genome to have eight or sometimes even more plasmids in the genome.

And sometimes those plasmids contain very interesting, relevant stuff, antimicrobial resistance genes, for example.

And when you just do short read sequencing, like a luminous sequencing of a genome like this, it can be very, very hard to tell whether a gene that you're interested in is on the chromosome or on a plasmid.

How much have you seen these assembly tools, whether it's long read first, whether it's unicycle or even spades, how much have they improved simply by virtue of the Oxford Nanopore biochemistry improving?

But what is it about bacterial genomes that unicycler is specifically tuned for?

So unicycler, it's a hybrid assembler, which means it takes both short reads and long reads to try to produce a nice complete assembly.

And it's what I've come to call a short read first assembler.

So it first will assemble your short reads into an assembly graph and And if you've seen a short read assembly graph, you know that they tend to be pretty tangled up because of any repeats in the genome, they collapse together.

And then unicycler uses long reads to try to untangle the graph.

So first it makes a tangled short read assembly graph, and then it scaffolds it out, hopefully to completion.

This approach was the ideal approach when unicycler was developed in 2017.

But how would you do it in your words? How would you describe how they work?

This episode is called "The Girl in the Fog," and you're listening to The Other World Patreon.

Hi, my name's Ryan Wick.

I live in, uh, Los Angeles, Koreatown.

I grew up in Phoenix, Arizona for the most part.

I like to make music and sew in my free time.

I've been making music for a little over 10 years, I think.

I do a label with me and my friends.