Hello, and thank you for listening to the MicroBinFeed podcast. Here, we will be
discussing topics in microbial bioinformatics. We hope that we can give you some
insights, tips, and tricks along the way. There is so much information we all
know from working in the field, but nobody writes it down. There is no manual,
and it's assumed you'll pick it up. We hope to fill in a few of these gaps. My
co-hosts are Dr. Nabil Ali Khan and Dr. Andrew Page. I am Dr. Lee Katz. Both
Andrew and Nabil work in the Quadram Institute in Norwich, UK, where they work
on microbes in food and the impact on human health. I work at Centers for
Disease Control and Prevention and am an adjunct member at the University of
Georgia in the U.S. Hello and welcome to the MicroBinFeed podcast. Today we're
joined by two special guests, Professor Ed Pfeil and Dr. Natasha Kutow. Dr. Ed
Pfeil is a professor of bacterial evolution at the University of Bath. His
interests include genomic evolution of pathogenic bacteria of both men and
animals. Early work is mostly on Staphylococcus aureus, where more recent work
includes gram-negatives such as Klebsiella pneumoniae, particularly in an AMR
and One Health perspective. But he's also worked on Borrelia, Burkholderia,
Wolbachia, Melissococcus, Rhinobacterium, Librio, Septococcus, Neisseria, E.
coli, Mycobacterium bartonella, amongst many, many others. And he also has an
interest in both bees and aquaculture. Dr. Natasha Kutow is a data scientist at
the Center of Genomic Pathogen Surveillance at the University of Oxford. Natasha
is a veterinary doctor and got her PhD in 2016. Her research focuses on the
molecular epidemiology, population genomics, and ecology of a broad range of
bacterial and viral pathogens of both animals and humans. She uses next
generation sequencing and bioinformatics to understand transmission of bacterial
and viral pathogens, and the emergence and spread of AMR between humans and
animals. She's worked on a range of organisms as well, including MRSA, Staph, E.
coli, Klebsiella, all of the entero-cocci and entero-mycobacterium, including
abscesses and TB, and flu, and also pigs. So welcome to you both. It's great to
have you on the show today. And we're talking about MLST, multi-locus sequence
typing. If you look at the original paper in PNAS from 1999, you will find the
third author is this Edward Feil fellow. So I decided we'll have Ed on to tell
us all of the dark secrets of MLST, what really happened 25 years ago. And
Natasha also as a user of MLST who mainly works in it from the genomic side
looking in. So it should be a fun, fun episode. So let's get started. I'll give
an easy thing that people might not understand. Most schemes, particularly, for
instance, like the Neisseria one, it's seven genes, right? Or most schemes have,
most schemes have seven genes or 10 genes or whatever. What's the decision
making for that number? OK, so for Neisseria meningitis, if you look at that
original 1998 paper, there are actually 11 genes used in that paper. There were
six, what we'd now consider to be classic MLST genes, which is the housekeeping
metabolic genes. And there were five more variables of outer membrane proteins
and so on. So there was there was half of one and half of the other. So it was
it was determined so. So to go back a step before we had MLST, we had a thing
called MLE, which was a gel based technique whereby you basically mush up, you
have a cell extract with enzymes still working and you run that extract through
a starch gel and you stain it with various chromogenic compounds that actually
change color to show you the position of how far the proteins got so you could
measure protein mobility. And you had various, what were called allozymes, OK?
So that was the and that was a really, really that was the first time we could
assay molecular variation in population. So that's the technique that went back
to the sort of late 60s. So and it was an absolute nightmare to do. So but there
was an MLE scheme for Neisseria meningitis, there was one person in the world
that could really do it. And that and she's still there. She's still in Oslo. So
that was Dominique Cogon. So the way it worked was that before MLST came,
everybody basically had to send their strains off to Dominique in Oslo and she
would run them on the starch gel because it wasn't a portable technique. So the
only way you could score an allele was to run it side by side with with other
strains which you knew that the allen were for so you could compare them
directly on the same gel. OK, so there was a and there was about 15 loci used in
that. I think it's about 15 loci used in that scheme. So when MLST came along,
and that was quite an established scheme, so you had the various different
lineages in the different steroid groups A, B and C. And when MLST came along,
the idea was that you should that it was necessary to recover the same lineages
that had been talked about before and defined by MLE, OK? So those those five
that were more variable loci, the non-MLST genes, they were thrown out because
actually they weren't bringing anything to the party. It was found that just
with those six genes, you could more or less recapitulate the MLE groups, but
not quite. So later on, there was a seventh, which was just put in just to
delineate, just to separate one group that couldn't be couldn't be distinguished
on the basis of those six genes. So it's calibrated entirely to fit with the
existing MLE genes and to have no more expense and no more resources than that.
So that that was the minimum amount of genes that you could have to basically
get what you got from the MLE. So that's why seven genes were both chosen. But
it worked as a reasonable sort of sweet spot, as it turned out for most of the
species. So it's just the tradeoff of expense versus resolution. We should
probably step back. I forgot to ask you, what is multi-locus sequence typing?
For those, for anyone out there who doesn't know, but someone might not know.
Having explained the multi-locus enzyme electrophoresis, so it's essentially it
was a typing method where you could define strains on the basis of not complete
sequences, partial sequences, up to about 500 base pairs, which was defined by
the sequencing, because that's that's how much you could get with with
sequencing runs going in opposite directions, how much clean sequence you can
get with with sequencing those strands. So it was a way by which you could index
variation, nucleotide variation in seven genes and a small number of
housekeeping, boring housekeeping genes, which were assumed to represent the
kind of underlying evolutionary relatedness, the phylogeny of the of the
population. So they're picked to be boring, they were picked to be under
purifying selection to minimise any confounding effect of diversifying selection
or recombination of those sorts of things. So and that that was kind of a novel
idea, which was and the novelty is basically demonstrated by the fact that in
the first scheme, Mark Atman insisted on using the more commonly used, highly
diverse genes, which were which were subsequently sort of thrown up. The other
really novel thing about it was that it uses new fancy shiny thing called the
Internet. And it was the first time because it sequence data is digital. It was
the first time that anybody had really put epidemiological databases up in a way
that anybody anywhere could actually. Type that follow this methodology, there
being a hospital in Australia or America or wherever, and immediately be able to
compare it with that database to see where that fitted in with the whole big
picture. And that is absolutely revolutionary because this was, you know, late
90s, the early days of the Internet, people just hadn't really sort of caught on
to how it could be used for epidemiology before. OK, and I should mention that
when you take each of these, so you have these set of loci, some some number,
and you take the sequence of that fragment and you're measuring the distances,
the differences between the number of differences between those loci for each of
your strains. And if the sequence is identical, then you give it the same allele
number. It's not counted. If it differs by even one base, it counts as a
difference. Is that correct, Ed? Yeah, that's correct. Yes. So I didn't explain
that. So so you end up with an allelic profile, so an allele number for each of
the gene loci that you've sequenced, so like a telephone number. And then that
unique telephone number itself gets a number, which is the sequence type, the
ST. And there was a reason why it was nonparametric in the sense that it didn't
matter whether an allele differed at one base or 100 bases in that it was it was
just, I mean, it made the whole thing easier to analyse when you just have like
a string of integers. But actually, there was also a recognition even very early
on that you could have recombination affecting an allele, which may introduce
one base, it may introduce 10 bases, it may introduce 20 bases. So in other
words, the number of bases by which two alleles of a given locus differed didn't
necessarily tell you anything.  thing about the number of evolutionary events
that have passed to explain those difference. So it was just either they're the
same or they're different. It wasn't, this is, these alleles are really
different from each other or these alleles are quite different. It was just
different or the same. Yep. And I will point out that you must recover the
allele sequence for every loci to have a valid sequence type. A few people have
come to me with missing things, missing profiles with missing alleles and
saying, what do I do with this? It's like, you can't do anything. It just didn't
work. No, that's right. All right. Natasha, do you have any, anything to add on
understanding MLST, particularly for people who come from a genomics background?
What's your take on it? I mean, I think for me when I was, you know, doing my
PhD and I, I started my PhD in 2011 and I didn't have access to whole genome
sequencing. And so for me, MLST was, was very useful because it was, it was a
way that I could, you know, communicate with other microbiologists or
epidemiologists and I could link my data into what's, you know, out there on the
internet, like, like I was saying. And so I think, you know, and I had to review
all these older techniques that were used for, for epidemiology. And I could
clearly see what was the benefit of, of using MLST and, and how, you know, how
you could kind of put a name on something or put a number on something and you
could speak to other people about, you know, E. coli SC131 would know what, what
that was and, and how, you know, pathogenic or dangerous a certain lineage could
be. And so I think, I mean, for me it was very useful when I started with my
master's and then with my PhD and it's, it still is very useful, very useful
typing technique to be able to, to speak to others about a certain lineage and
how dangerous or not it can be. So yeah, I'm curious about that from the epi
perspective, because when I started, I remember going to conferences and seeing
a lot of people say MLST was no good because it didn't have the discriminatory
power of PFGE. And PFGE was the way to go, particularly at an epi sense. What
is, what is PFGE? Why would you choose one over the other? And maybe dare say,
which is better? So PFGE is Pulse Field Gel Electrophoresis, which is another
band based method. So that immediately puts it in the same category as the, in a
sense as the original, you know, MLE, where you had to compare samples directly
on a gel. It was very much standard. It's also, you have to have quite a
dedicated lab to do it. It, it was, it got quite sophisticated in, in how far it
got digitalised. So there was, there was, there was pretty good software that
enabled you to sort of scan these gels and, and define your strains on the, on
the basis of the, of the mobility patterns you saw. But it was still not
digital, like sequence data is digital. It, it took quite a long time to, to
fizzle out because there was a lot of investment, particularly in the States, I
think, in, in, in PulseNet, which was all based on Pulse Fields. And, you know,
once you put that investment and, and people know what they're talking about and
everything sort of seems to work, then it's very hard to sort of turn that ship
around. Whether it's better than MLSC, whether it, it, it, it caused more higher
resolution, I think it actually depends on the species. So, because, because it
works by, it's a restriction, it's a restriction fragment digest, right? So
you're just, you're just picking up variations in the presence, absence of
particular restriction sites in your, in, in the genome. So in that sense, in a
sense, it's, you don't know, it has the big disadvantage in that you, you, you
can't go in and actually look at the, what's causing the variation very easily.
So you don't know exactly what's going on in terms of the genetics that cause
the diff, the variation you're seeing. But at the same, but it did, it was a
genome-wide technique. So it did detect changes in accessory, you know, we
didn't really know that there was such a thing as an accessory genome in the, in
the 90s, but it was detecting parts of the genome, it was reaching parts of the
genome that MLSC wasn't. So whether it's better or not kind of depended on the
species. You know, if you've got quite a stable genome, then MLSC was fine. If
you've got a lot of variation in the accessory genome, then PFGE was picking
that up and MLSC wasn't. So it swings around about to, to a degree there, but
really the, the, the, in the end, the advantage of MLSC was it's because
sequence data is digital and you can put it on, it's much easier to store and
much easier to compare. And that actually quite quickly became much easier to do
than PulseField, outside those big dedicated labs that were just like big
factories for doing it. Natasha, what about you? Did you encounter PFGE early
on? I did a lot of, yeah, I did a lot of PFGE. As I said, I, I, I didn't have
access to whole genome sequencing at a time. And what we had available was PFGE.
We would start off with, with MLSC, of course, and then we would, you know,
carry on with, with PFG, especially if we were trying, you know, like I, I
didn't mention this, but I, during my PhD, I was trying to find strains, strains
that was similar between animals and humans, all sorts of animals, I have to
say, not only pigs. And so it was, to me, I had, I needed that resolution that
PFGE could give me that, that MLSC could not, because I could find, for example,
ST398 in humans and in animals, but I really wanted to, you know, be sure that I
had enough resolution to say, okay, this is probably, there was a transmission
event here, or there wasn't a transmission event where, you know, there was
dissemination of, of this certain strain or there was not. And I think at the, I
mean, at least I was not working at a human hospital at the time, but the idea
that I have is that when, when it came to hospital outbreaks, they would be
using PFGE, like we kind of use now SNP analysis to define, you know, what, what
strains were belong to a cluster or, or to an outbreak and what, which strains
were, did not belong to, to an outbreak. I think at least for MRSA, that was the
level of resolution that was needed and at the hospital level, although like Ed
said, you know, you really needed very dedicated people to, to do this. And of
course there were also other, there were also other options like MLVA or what
was the name of the, of the rep PCR, I think it was for E. coli. So yeah, you
had other options, but at least when I, when I was, I was using PFGE to, to have
better resolution than, than MLST for MRSA. I've been really enjoying the
conversation because I don't know if you know this, but I, I started off in
meningitis and so my whole thesis was on nice Syria and everything. And so a
bunch of your stories was just, were just giving me flashbacks. Like I, and I
was in the, I was actually in the CDC meningitis reference lab for a year. So I
remember names like Dominique. I remember looking at the software that Keith
Jolly came by to set up stars and everybody doing MLST. So it was very
interesting. And I don't know if I have a specific question, but it's just been
really good hearing it all. I noticed that you don't really have nice Syria in
your biography. Have you, have you kind of moved on from that? Oh, well, well,
yeah, I have, I haven't, I haven't worked on. So my PhD was on nice Syria,
meningitis and gonorrhea. So, so my PhD was, was to sequence ADK and RecA in
meningitis and gonococcus, each in seven genes, each in seven strains. So there
was a site and gene sequences and that was my, because he had to clone it and
everything, but they, they were the, they were the, they were the, the sequences
that went into the MLS, the final MLST schemes. So that was my, so those two
loci were my contribution, ADK and RecA. If I remember that paper had like 107
genes. It had 107 genes. Yeah. So I didn't do all those. It's because I'd
sequenced those because they weren't, you know, right before genome sequences,
you had to clone the gene to get the sequence. And that took a year, a year
anyway. A lot of work. So yeah, yeah, yeah, yeah. Thanks for the walk down
memory lane. Yeah. Now people complain if the Illumina run takes more than three
days. I actually have a question for Ed, if, if that's okay. Yeah. I was going
to mention eBursts because you were the developer of eBursts, right? So why, why
was there a need to do eBursts? Well, actually in a sense, this comes back to
what we were talking about on the other podcast, when it comes to, to figures.
Before, so the standard way of visualising the, the MLST data originally was
using UPGMA dendrograms.  And I distinctly remember my supervisor, Brian Spratt,
he kept a magnifying glass in his office drawer so he could look at the
dendrograms that came out, even with like a hundred strains, they were really,
really hard to actually read and look at. And he literally had the magnifying
glass in his desk so he could read all the labels and so on. So I just thought
there'd be a better way of drawing it. And Brian had sent me this project which
involved, it actually came from a paper from Gutmann and Dichowsen where they,
on E. coli, where they showed that you could actually, if you took really,
really similar sequences in E. coli, you could actually directly just score
mutation events and recombination events because you haven't got all, you
haven't got to dissect out all the subsequent events. So this is a trick which
has stayed with me to this day, if you look at really, really closely related
stuff, everything's a lot easier. So he said, look at the MLST data and see if
you can figure out whether alleles differing by recombination or mutation
according to how different they are and try and do something quantitative with
that. And that came from that paper. And then as a process of doing that, it was
really obvious for the Staph aureus data and for the Noceum meningitis data that
actually you had this model where you had a central genotype and then you had
spokes coming out from that central genotype which differed by one locus. So
one, I invented the SLV during that time and that became the one. I said, well,
that's what's happening. So why don't we just draw it like that rather than this
funny dendrogram where you've actually got the founder, the ancestor actually at
the same level as all the descendants. Let's just put that in the middle and
have all the sense and not worry particularly at that point about how to connect
all those different groups up. Let's just go with that model and draw it. It's a
really simple idea, but it kind of just, I guess, got people thinking different
ways of visualizing, thinking in circles and this different way of looking at
visualizing the data, getting away from that classic dendrogram tree shape which
was actually not only almost impossible to look at but actually quite misleading
in many ways because you didn't have that model of clonal expansion implicitly
in it. So that's what kind of led me to Ebers. And the actual nuts and bolts of
it are absolutely incredibly simple. I mean, there's nothing clever there at
all. Oh, no, Ebers is great. I wouldn't put yourself down on that. So if you
look at the original paper, which I think is in JBEC 2004, and you look at the
first figure, that is precisely what Ed is showing. There's this head-to-head of
this very fairly, it's a simple dendrogram, but it's actually quite ugly to look
at. And then you have- It's really ugly, and it doesn't tell you about how the
thing's evolved, right? It hasn't got that clonal radiation thing going on. And
then Ed's very cleverly just slapped the actual Ebers figure equivalent in the
middle of it just to show like, hey, see this trash? Forget it, this is the new
hotness. And then you look at the other couple of figures, then there's these
beautiful constellations of clonal complexes. And I think that's the magic,
right? As you were saying, one loves a good figure. So even with data sets, they
look quite gorgeous. So is it you to blame for the use of minimum spanning
trees? Well, I think so. So I did actually contact Bionumerics about this idea
at some point. I didn't hear anything from the minimum spanning trees appeared
in their next version of their software. But I mean, and essentially, I mean,
there are some differences, but the nuts and bolts of basically identifying the
center of gravity of these clonal complexes on the basis that they define the
maximum number of near neighbors. That was the real trick. So I was always a
little bit frustrated that they called it a minimum spanning tree because
there's lots of different, like hundreds of thousands of different solutions,
minimal ways by which you can connect all these genotypes. But it's not until
you actually apply this model of having a central founding genotype defined in
terms of the maximum number of neighbors that you can actually pick one solution
that kind of makes sense. You still have to be a bit arbitrary if there's
details where there might be ties in how you rank things and stuff. But that
basic nuts and bolts, it was the same in the minimum spanning tree, which Joao
and his colleagues in Lisbon recognized very early, and they came up with
PhiloViz, which is brilliant, which was actually definitely an improvement on
the original universe. What would you have called minimum spanning trees if you
could have named them instead since you had an issue with them? So, well, I
don't know, really. I never really thought about it. I mean, it was a minimal
spanning tree, but it was misleading in that it was one possible minimal
spanning tree out of lots of different solutions, right? And so, oh, I don't
know. Eberst? Call it a phylogram. Yeah, man. But I've moved on. It's all right.
Well, people still use things like minimum spanning trees to this day for even
CGMLST data. Yeah, yeah, no, absolutely, absolutely. And I do like to think at
some level that I got people thinking in terms of circles rather than
dendrograms, which is a nice contribution, I think. I was wondering, like, as a
user of MLST, and I've recently been involved in getting huge datasets and doing
MLST, so, yeah, getting the typing for numerous different species, do you think
the quality of the genomes, the quality of the MLST schemes that derives from
that original paper were as good as the ones that you helped develop? Because I
kind of have the idea that, at least for some species, like pseudomonas and
Acinetobacter, that maybe because sequencing was not so available at the time
that they maybe didn't select the best genes. I think that's completely fair.
There's a good example. It's in the pneumococcus of the gene, which is just
absolutely, I think it may even be two copies of it in some genomes or
something. There's something really fundamentally wrong with it. Yeah, I mean,
and again, those genes were picked before there was a genome sequence available.
So they were picked on the basis of them being most likely to be, as I said,
under boring purifying selection, central metabolism genes, nothing particularly
interesting going on in terms of their evolution, but we didn't know how well
they were gonna pan out. And I'm sure that in almost all of the schemes, there
were some genes which have, some choices were better than others. I'm sure you
could have optimized them, but that's, we were working in the dark a little bit.
I wanna ask a question that follows on from that, but it's a bit more fuzzy. So
I read, I think it was you in a review with Martin Maiden that I was reading
recently, and it talks about how, about clonality in bacterial species. And when
you think about it, MLST is assuming that the population will fall into these
discrete clones, that there are genes that actually will be boring and sort of
stiffen and reflect that. Do you feel that, but I do know from certain
organisms, so E. coli at a point, if you take the ML, if you look at the
original MLST scheme in that paper, they show that it kind of works, but it
doesn't. Like there's a point where it starts to break down the certain parts of
E. coli where that assumption doesn't seem to hold, at least in the data
presented in that paper. And I'm curious, how do you feel about clonality in a
bacterial species today? Is that a given? Do you expect that to happen? Or is it
because we started looking at pathogenic clones first, and maybe fooled
ourselves a bit that that was- Yeah, well, I think this was a realization that
came straight away, actually. I mean, there was a big debate around the sort of
mid-90s. There was a big debate about how much recombination is going on in
bacterial populations, because all the MLE work from the big American labs, Bob
Salander and Howard Ockman and that crowd, they're always saying, bacteria don't
recombine. It's all clonal. Everything falls into these nice little groups. That
was the dogma. When sequence data came along, even before MLST, that dogma began
to be chipped away at a little bit. And there was quite a heated debate about
where bacterial populations sat between the two extremes. On one hand, you had
clonality, no recombination, everything fitting into these nice groups. And on
the other hand, you have panmixia, which isn't the word you hear so much these
days, but that was the idea that everything's so mixed up that there's
absolutely no lineages at all.  So, this came to a head with the Maynard Smith
paper, How Clonal is Bacteria, where they basically took the data that was
available at that point, the MLE data, and it was MLE data, and basically,
because this was before MLSC, and basically categorised different types of
populations. So, from panmyxia, which was gonorrhoea, the gonococcus at the time
was, in that paper, was described as having no clonal lineages at all,
everything was just completely equally distant from everything else, and the
alleles were in linkage equilibrium, so everything was just completely randomly
shuffled to clonal species. And then, in the middle somewhere, you had Neisseria
meningitis, which was described as this epidemic structure, which is basically
referred to the fact that you had a soup of recombining things, where alleles
are just flowing backwards and forwards, but on top of that, you had these
croutons of clonal complexes, which were the virulent clonal complexes, which we
were mostly focused on. So, that was the sort of first realisation that you
could actually have a bit of both, you could have clonality, and you could have
a lot of recombination going on at the same time. And then you had to explain
that kind of paradox in terms of, what are those, is it all just oversampling
particular clonal lineages partly, is it because those clonal lineages, the
particular combinations of alleles are adaptive, and this is a selective thing,
keeping those things together, probably partly as well, or they could just be
hitchhiking on one particularly favourable gene. So, there were nuances entered
into that whole population structure from very, very early on, because it was
all about those early, that early work was all about answering the question of
how much recombination is going on. So, there is confusion about when people say
it's a clonal population, they can infer that it's not recombining very much,
and that may be the case, but it's not necessarily the case. So, there's ways by
which we've known since those days where you could have clones and still have
quite a recombining population. So, no, I mean, even from those, I think the
closing words of my PhD thesis was something along the lines of, there's no
pattern in bacterial population structures. So, I haven't been surprised at any,
anything that's come since then, because I had no expectations of what anything
would look like. Saphoris is a beautifully well-behaved, lovely organism,
something like Klebsiella or some of the Vibrios are just all over the place,
you can't tell what's what, something like Burkholderia, you have not so much
sequence variation, but you have a lot of allele shuffling and there's
everything in between. So, yeah, and we've known that since day one, really.
Natasha, what about you? How do you feel about clonality? I'm also curious if
you feel that it's, is it useful to think, to teach it that way? Talk about
clones all the time. And I'm also curious, based on the organisms that you've
both worked on, other organisms you worked on, which, which would you say are
more clonal, which are panmixia, which are somewhere in between, just as a
schoolyard question for people to remember. So it's interesting to know what to
expect. Yeah, I mean, I started realising that maybe, you know, actually, I
realised that MLST was not, was not a good choice for me, when I started working
with staff through the intermediate. First of all, because the MLST scheme that
had been developed for this particular pathogen only contained initially five
genes. And, and then it was later changed to seven, seven genes. But when I was
looking into the resistant population, there was a population structure, yes. So
I was mostly seeing certain clones, ST71, for example, was the main RSP clone at
the time when I was doing my PhD. But when I was looking into the susceptible
population, I could rarely see the same MLST. I mean, any, you know, any
sequence, any strain that I was typing that I was doing MLST, I would find a new
ST. So basically, there was no population structure, they were all different,
maybe, like, like, we talked about before the MLST scheme was not, was not the
best. And definitely in the beginning was not with just five genes. So that was
when I first, you know, was first confronted with this, you know, idea of
population structure and clonality, something that I heard, I had heard a lot on
staff areas. And for me, I think MRSA is, you know, is the best and the
classical example of, of, like, Ed said, well behaved pathogen that, that
expands through, you know, through basically vertical transmission. And there's,
there's this population structure and these clones that are, that are important.
I think we, we, we started, I mean, I wasn't there, like, like Ed was in the
beginning. But for me, when, when I was reading this, this papers and hearing
about clones, I think it made sense at the time, because, you know, and it still
makes sense today, because you can define these clones based on the fact that
there are more frequent in, in the resistant population, for example, if you,
you know, if you take a sample, and you play those, the strains, and then you
sequence them, you realize that there's a few, a few lineages that appear more
often than others. So there's definitely a higher risk of these certain lineages
acquiring certain mobile genetic elements, plasmids or integrons or, or
whatever. And then other lineages are more prone to acquire virulence. We don't
really know why this happens most in most cases, but there's, you know, there
was, there's definitely, I think it's definitely useful to talk about clones,
and they exist, because we've seen a certain pattern in the past, and we see a
certain pattern still today. Something Natasha touched on earlier when she said,
you know, it was used for E. coli 131, ST 131, everyone knows what that is now.
And that's that, that, that, that I think is the, the legacy of MLST for all its
flaws. There's not, there's not many examples I know of where in the genomics
era, key sort of conclusions from the MLST data have been proved completely
wrong, which is pretty amazing when you think, you know, the genomes only had
seven genes in those days. So obviously, they've been refined and tweaked. But I
think the lasting legacy of having the nomenclature for those, those, those key
lineages, clones, whatever you want to call them, is it will be the gift that
MLST leaves us really. And that, I mean, I guess that that will be set in stone
forever now, because those lineages are real, we know they're real. You can say,
well, let's split ST 131 up into two different things, or let's combine it with
something else. But the lumpers and the splitters thing is always going to
happen. So, so, so good for MLST. Well, it's not, it's not dead. I don't think
it's finished. No, no, no, no, no, no. Well, no, I mean, the database is
certainly on. Yeah, I mean, it's, it's, do people actually just sequence seven
genes these days and not the whole genome? I think so. I think people still,
people will still do it in reference labs out there. Some people do not. Okay. I
thought it's actually easier to do whole genomes these days. For most, for some
of us, yes. But for some people, it's still difficult to get access to whole
genome platforms. And so MLST is the standard or MLST is what the provincial
labs are capable of doing. And that's what they stick with. So people, I, you
know, remember Keith telling me that people still submit Sanger traces to
PubMLST. Right, right. You know, so it's still there. And, and the, and the fact
is, is, is if proprietary companies just disappear tomorrow, you can always go
back to it. It's a free and open, it's free and open software, free and open
science. Just to do it, the primer sequences are in the paper, go for it. Yeah,
yeah, that's true. So yeah, I think, I think it's not quite, quite finished yet,
but for the most part, I think it's, yeah, it definitely has. Most of the major
labs have moved to straight genomics now. There was, there was a strange time
right at the beginning of the sort of genomics era and that there were the next
generation platforms where people are actually sequencing whole genome. I saw a
couple of papers where people would sequence whole genome, but just report the
MLST, just to do MLST. It's like, what about the other 2000 genes? Anyway. No, I
still see that. And I see people using genomic data just to report the serology.
Right. Things like that. It's like, okay, it's a bit, it's a bit overkill, but
all right, fine. Whatever works. I mean, it's just, it's yeah, you put any
organism in it, it's the same pipeline. So I guess it's more efficient in the
end. All right. Well, that's all the time we have for today. I want to thank
both of our guests, Ed and Natasha, for joining us again, and we'll see you next
time on the MicroBinfee podcast. Thank you so much for listening to us at home.
If you like this podcast, please subscribe and rate us on iTunes, Spotify,
SoundCloud, or the platform of your choice. Follow us on Twitter at MicroBinfee.
If you don't like this podcast, please don't do anything. This podcast was
recorded by the Microbial Bioinformatics Group. The opinions expressed here are
our own and do not necessarily reflect the views of CDC or the Quadram
Institute.