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 US. Welcome to the MicroBinFeed podcast. Today we have a special
panel discussion. We're at the seventh microbial bioinformatics hackathon, and
this time it's got a special focus on AMR, and it's being organized at JPI, AMR,
PHAGE, and CLIMBIGDATA. We have a brilliant panel here, and I'm going to ask
each person to introduce themselves. Mark? I'm Mark Pallon. I'm a professor of
microbial genomics at the University of East Anglia and a research group leader
at the Quadram Institute of Bioscience. In a former life, I was a consultant
medical microbiologist. I'm medically trained, but I've seen the light and
become an academic in recent years. I dabble in bioinformatics from time to
time. Hi, I'm Phil McGuire. I'm about to start an assistant professorship at
Dalhousie on the East Coast of Canada in computer science and epidemiology.
Yeah, I do a lot of work on AMR, AMR metagenomics, shitting on MAG methods for
plasmids, and with CARD and SARS-CoV-2, so just kind of all over kind of AMR
bioinformatics and infectious disease bioinformatics. Hi, I'm Anthony Underwood.
I work in the Centre for Genomic Patterns and Surveillance in David Anderson's
team. I get most excited about actually implementing bioinformatics and actually
seeing it in action. Recently, I've been involved in a global health research
unit project, which is about implementing global surveillance for AMR based on
WGS. That's been very interesting seeing how when you have groups who are
perhaps novel to the field, how do they go about implementing it and what are
the challenges they face when they're doing it from the ground up? So my name is
Clement. So I'm working in a hospital in Qatar at the moment, although maybe I
may be in a transition very soon. But yeah, I'm a research scientist, so I
basically do whole genome sequencing for the AMR pathogens collected from the
patient, both bacteria and also fungus. So I am a molecular scientist in a
clinical lab and working with the physician and the laboratory staff in the lab.
So I help building the molecular lab for infectious disease diagnosis and also
involved in setting up some basic pipeline and workflow for whole genome data
analysis. AMR is important in the Middle East because I think very few people
are actually looking at the bugs in that part of the world. And I think in that
region, people are getting, they are more concerned about the spread and also
the transmission of the MDL pathogens in the Middle East. I think there's lots
of novel discovery during the past few years when I was working in that part of
the world. And actually not many hospitals are using whole genome sequencing to
do the detection and also diagnosis. So they basically still rely on the
conventional techniques and some of the quick genotyping tests available in the
commercial platform to detect whether the bugs carry MDM or KBC or just oxa
instead of looking in depth of the gene, the presence or absence of particular
genes in the bacteria. Awesome. Okay. So my first question really is, what are
the major challenges for getting AMR genomics into the clinic? So, you know, we
can all see medical microbiologists really want to know immediately what is the
AMR profile of a, of a pathogen and we know sequencing can get those results
pretty quick, but you know, where is the disconnect? What do we need to do to
push it over the line and get it a bit closer, Anthony? I think one of the
issues is that there's still some scepticism that for genomics at least that you
can get a really trustworthy result from the genomics data. So for some bugs
like TB, there's a growing acceptance that you can get faster, better results
using genomics, but for other bugs, Pseudomonas acinetobacter, there's quite a
lot of discordance between the phenotypic results and the genomic results. And
so I think for some bugs, the path to actually seeing it into the clinic is much
more straightforward given cost and time, I suppose, has been one of the factors
which are very important, but for other bugs, I still think it's a long way to
go and it'll be some time before they can be implemented in the clinical setting
and will probably remain in the domain of the surveillance setting for the time
being. So what is the lowest hanging fruit in Europe? In terms of bacterial
species? Yeah. Oh, well, like I said, TB is the obvious one because I mean, it's
implemented already in the UK, other species, Salmonella, there's a pretty good
congruence. Some of the enterobacteriaceae, Klebsiella is not bad. I think
there's some pretty good one, pretty good concordances with those data. Mark?
Well, I was on a learning curve a few years ago when we were doing some work on
ancient TB. We were looking at some 200 year old genomes from some mummies. And
it was only after a certain amount of time, scratching his head, that the
bioinformatician suddenly realized that what he was looking at was a mixture of
two genotypes in the same sample. And I said to him, I said, I don't, I don't,
that doesn't tie in with my clinical understanding. You don't catch TB twice.
It's very unusual. What's going on here? The thing was that his initial pipeline
that he used had actually excluded many of the things that were in front of him
because it was looking for SNP calling and it was excluding anything that didn't
meet a certain criteria, like 70% representation. In fact, he had two different
genotypes mixed together, just by chance, roughly 50, 50, 50% each. And that
made it clear to me, when we looked at the literature, it was clear that
actually there is quite a lot of evidence of mixed infection with TB. And in
some areas where TB is highest prevalence in KwaZulu-Natal, I think it was one
in four patients has evidence of mixed infection. What we found when we were
looking at these samples from 200 years ago in Hungary, when probably in Hungary
at that time, the instance of TB was similar to that in KwaZulu-Natal, very,
very high, was that, yeah, the majority of the samples had mixed cultures. And
when you're trying to predict AMR, that becomes an issue because if you have a
strain that is sensitive, mixing with a strain that's resistant, you've got to
have sophisticated enough pipelines that can actually take that on board and
say, give you an honest report of what's going on. The problem is that, you
know, what is clinically interesting and actionable, if 95% of what you see is
sensitive, but 5% is resistant, and then you give the antibiotic to which that
5% is resistant, you'll probably end up with 100% resistant as your majority
culture. So even minority populations are important in this setting, but it's
certainly not well-defined. The other issue that became apparent to me when we
were looking at that issue, we were using metagenomics. What was happening at
the time is that people were taking single colonies in lab and following it up.
So they're actually purging the existing biological variation out of the sample
and just following up one thing. According to Conor yesterday, he said that,
well, nowadays, people will probably take a sweep to try and keep that variety
alive. But this is something, if you're going to future-proof your procedure so
that in the age of metagenomics, where you don't want to actually have to
culture anything, you just want to take the sample, extract the sequences and
call out the genotypes, including resistance genotypes, from the metagenomes,
this is something you have to take into account because there's a good chance
you will see a mixture of things. The other thing, the other one last point I'd
make is that there are two reasons why you get this kind of effect. One is
within host evolution. So you may have one strain, which part of which has
evolved to become resistant. So that part of the population is resistant, the
other part isn't, it's all the same genotype. Contrary to what I said earlier,
people do get infected twice with TB. And when you think about it in an area of
high prevalence, where you might be coming into contact with TB every day in
crowded situations, why would you not get infected twice? But in those
situations, you may have two completely different genotypes and you have to pull
those apart. The worst case scenario is where someone gets infected twice with
distinct but very closely related genotypes, where then it becomes even harder
to pull them apart. So Mark, do we know, is this like a cloud of infection
people get infected with or would it be distinct different infection events? So
when someone gets infected, do they get infected with a single cell and the
whole infection goes through a single cell bottleneck? Or let's say they got
infected with 10 cells, five of those cells are one genotype, five of those
cells are the other genotype. I don't think we know what is, you know, these
kind of.  situations with HIV, you know, that things go through a bottleneck,
but it doesn't actually necessarily purge all the variation out of it. Sometimes
it does, sometimes it doesn't. With TB, I just don't think anyone would know. I
don't think it's been looked at enough to know quite how often you do get mixed
infections passed on in a single transmission event. So it's fascinating,
actually, the population dynamics. I think one of the barriers for genomics,
especially metagenomics in the clinic, is it's the amount of data they actually
do produce, and it's the interpretation of that data. It's a lot, and I've
personally been working at Sunnybrook trying to help them get accreditation via
the medical microbiology kind of accreditation organizations for their use of
pathogen genomics. And frankly, the accreditation agencies don't really, that
I've dealt with, don't seem to have any idea what they need to do to accredit
use of genomics. So that's a huge, huge logistic and legal barrier to getting a
marginomics really kind of more widely used. Because, yeah, how do you deal with
the fact that, yeah, okay, compared to our simple PCR test that we did on the
TB, and we found there was, everyone had a single infection. We found this one
gene, so it's resistant to, this antibiotic is great. But when you find, oh no,
it's a mixed infection, and one of the strains is resistant to this set of
antibiotics, and the other is probably resistant to this set of antibiotics, and
now you need to use your medical judgment to work out what you can do with that.
Which is where I think the other kind of major component comes in, and some of
that, you know, trying to do that with phage and harmonization is kind of
standardized reporting of the results. Like, how do we actually report those
genomics and metagenomic results to clinicians so they can make meaningful
decisions without kind of doing our bioinformatics thing of like, here's all the
data, work it out. You know, it's like, oh, but this could matter, but you know,
maybe it won't. So as I think some of Anna Christensen's work a few years back
was really cool, especially for English for TV, it was focused on, which was
kind of using evidence-based approaches and speaking to clinicians kind of in
this iterative process to develop, you know, a meaningful report. So yeah,
accreditation reporting, and I think those kind of tie together. And also just
genomics and metagenomics kind of not being penalized for being more informative
than more simple tests, and working out how to kind of balance that. Yeah, I
guess too much information can actually overwhelm people. And Anna's very famous
talk, I think, was at ASM NGS. I think it was one of the best talks ever that
I've seen. And it was very, very clear, and it made it very straightforward and
simple to see, you know, this is the information that clinicians need versus the
information that, you know, me as a biomedicine practitioner would want to
convey. And I'd want to tell people, you know, exactly where the symptoms are
and exactly which allele it is, but it doesn't impact clinical, you know,
decision-making at all. You know, it just overwhelms people and they ignore it.
So I'm going to go to Mark Pollan next. Yeah, I'll just amplify that. I mean,
the context is everything. And what you can tell through metagenomics, and then
through the bioinformatic analysis of metagenomes, may or may not be actionable
in the clinical setting. And it's very easy to get carried away and say, oh, we
can show these interesting things, and then find that actually nobody's going to
take any action. Basically, when clinicians use antibiotics, they're being
defensive in their practice. They're always going to go for the broad-spectrum
antibiotic because they know then that they've covered all the options. So for
you to say, well, we've got this organism and it's sensitive to this, you know,
narrow-spectrum antibiotic, then they're probably not going to take any notice
of you. What we did a couple of years ago was we looked at the intensive care
unit microbiome, and we looked at the gut microbiome of patients there. And this
was another example where my assumptions were challenged. I assume that if you,
these patients all get given antibiotics, 95% of them get given antibiotics when
they go in ICU, you'd expect there to be a little bit of resistance in the gut.
What we found was that the gut microbiome was taken over, overwhelmed by usually
a single clone of an antibiotic-resistant organism that basically became 95% or
80, 95% of that metagenome. And so you'd think, well, you know, you've detected
this highly resistant enterococcus in this metagenome, surely this is clinically
actionable, or you've detected a, you know, multidrug E. coli. But when we
delved into the literature, I mean, obviously, you could do the same thing by
taking, say, people take perianal swabs and then plate them out and say, oh,
yeah, we've got multidrug-resistant E. coli in this person. But the evidence to
show that that is a cost-effective approach, even using conventional
microbiology to go and deliberately screen for antimicrobial resistance in
carriage in those patients is not well established. So we kind of thought, oh,
we're leaping ahead into doing it by metagenomics. But when you look back, even
doing it by conventional approaches, it wasn't well established. So just because
you can do a thing using metagenomics and bioinformatics doesn't mean that it'll
actually be useful in the clinical setting or adopted. You have to be very
careful to question all the assumptions in that whole pipeline from doing the
analysis to actually being deployed somewhere. Just briefly a follow-up, because
you comment similar situation. When we do WGS, we will generate lots of
information, but not all the information are relevant to the physician. Usually
in their mind, they are looking for a specific target for making clinical
decision. And I think NGS data usually are more for retrospective. It's rarely
been used as like a prospective study. I mean, it's not quick enough comparing
to those simple PCR tests. But definitely progenome sequencing would offer much
more data in a short period of time. If we can generate the data within a few
days, definitely some of the WGS data will be supplementary to the follow-up or
the treatment, the patient in that sense. But of course, WGS data sometimes
would give us some additional data. Say for example, in two cases that I
encountered, we sequenced some MDR bacteria during the screening or from the
patient. We detected MCR1 gene in two bacteria, which is to their surprise,
because if you need to confirm the bacteria and high MIC2 was callisthenic, you
have to do some sort of specific microdilution test in which not many hospitals
are equipped to do the microdilution test. But then when we detected the MCR1
gene in those spots, in addition to the CRE, then the whole scenario, the
context changed because the physician didn't know that the bacteria actually is
callisthenic. The system is like even hyper kind of things. Then it's alert.
They're more sensitive in that way. So there's some special cases like the WGS
data just getting just often more for the physician. But in other scenarios,
probably the physician, they already have enough data on their hand based on
their phenotypes to make a clinical judgment, but they really wanted to have WGS
data just to confirm. So they would rather have it done. But I mean, it's a
selective process. We are not, as a molecular scientist, we are not making
judgment which bacteria to be sequenced. It's always their call. So they wanted
to look for something additional to look for evidence to confirm the initial
phenotypic data. But to me, as a scientist who analyzed those data, I found it
to, it's really hard for me to QA those phenotype data because the phenotype
data is an output from the phoenix or biotech and they are stored in say a
hospital information call center or whatever system the hospital is using. And
those platforms are not authorized. I have no authority to assess those data.
They can print out a hard copy and then give it to me. Okay. It's like, it's a
whole bunch of the MIC of this part, but like it's in PDF and things like that.
It's really tedious for me to type in, to translate those information in the
Excel sheet or whatever, because it is like an output PDF. So I don't know how,
so that's why in many, in a discussion in those metadata, I said like many of
the bio-sample we submitted to the, I would guess many of this bio-sample
submitted to the GenBank, the MIC are not submitted as well because like those
data are in a different domain which is not accessible either. It's like PDF.
How do you find student help to type all this data into Excel sheet or in the
bio-sample submitted form? It's impossible. So I'm just wondering from a
clinical point of view, do we need to go direct from sample in terms of speed in
order to be able to make this something which is viable? So for TB it's a
special case because it's a slow, but for other bugs, do we need to be able to
have a solution which is direct from sample rather than having to go through the
isolation and whole genome sequencing? So that was kind of the question from a
clinical perspective. And I guess from the bioinformatics point of view, it
still seems to me that that metagenomics approach is still a little way off
being ready for prime time. So I just wondered whether people had comments on
that perspective. You mean to go direct sequencing without a culture? You mean?
Yes. I mean, I think that, yeah, it's an interesting question because it's one
of the things that I've taken seriously. You think, oh, wouldn't it be great if
we could do this? We could cut out so much time and effort. And then you sit
back.  back and it will be interesting. I mean, there's a literature on it about
how often is action taken on the result of laboratory tests. And you have to be
sanguine about that. You think that everything you're doing is really important,
it's gonna make a big difference. But as clinicians will often just, they'll
say, well, we're gonna have to give the broadest spectrum antibiotics or we're
gonna have to be very careful. We're gonna start treating this patient even
though, first of all, they start treating the patient without having a specific
pathogen. And often they don't get a specific pathogen, but they say, well, we
think this patient is infected, we're gonna give them broad spectrum
antibiotics. When you do get a specific pathogen, they may or may not change. I
mean, but usually only in those cases they escalate upwards to using more
powerful, more broad spectrum agents. Very seldom do they go down. And even when
you tell them, oh yeah, we've got this multidrug resistant thing, how much
action is taken? And actually I was putting my hand up earlier for another
point, which I think should be made. When you're talking about the
bioinformatics of AMR, obviously the one key point is detecting AMR, detecting
the resistance genes, predicting phenotypes and so forth. But another key role
that we can use genome information for is in unpicking the AMR. So how is it
spreading around the hospital? How is it spreading around the community? We did
one study where we tracked a multidrug resistant acetobacter over a very
prolonged outbreak in a hospital. Part of controlling AMR is antibiotic
stewardship and appropriate use of antibiotics. But infection control measures
are just as important. And so knowing where the organism is, how it's spreading,
what we discovered in this study we did was that there was cryptic spread going
on in a burns theater, where patients who had burns were getting infected and
then they're getting their burns debrided in this operating theater. And that
was where the organism was traveling from one patient to another. But we were
able to show that because we could see that the genomes from successive patients
who'd been to that operating theater were identified. And so their whole genome
sequencing allowed us to control antimicrobial resistance in a way that you
might not anticipate by just thinking all you're doing is predicting the gene.
Predicting the genotypes and the epidemiology is really important. That's
simple. And with the intensive care unit study we did where we, there we did
look at metagenomes and we got, we were able to genotype the vancomycin was just
enterococci we're getting out. And we were able to show that in some of the
patients they had indistinguishable genotypes or one or two SNPs different. So
it was inconceivable. This was just by chance. These patients had acquired this
from a common, from each other or from a common source. And we went back to the
clinicians and we said, what about these patients then? And they said, ah, yeah,
they were in adjacent rooms in the same part of the ICU at the same time. So
what you're saying is that there was cryptic transmission of antibiotic
resistant organism between patients. We didn't screen for that. We wouldn't have
known about it. They were colonized, not infected, but nonetheless it shows that
our infection control measures are not good enough. Having this genome based
information and being able to say beyond all reasonable doubt that these things
are transmitted here in this specific context, it does inform behavior beyond
just treatment with antibiotics. Absolutely. And this kind of prospective
sequencing does help set the context for a lot of this stuff. I think
practically WGS can, we are able to sequence the bacteria without the culture,
but it also has to be in the context. I think we have to create a workflow that
justify for the diagnostic lab to do in such a way, because in a clinical
laboratory on a monthly basis, we are culturing almost 6,000 bacteria. So
depending on different contexts, like screening, some doctor, physician,
infectious disease division, they are asking to do screening, say swap from
skin. And then the 6,000 culture came from skin swap, urine, blood cultures,
whatever, from different kinds of collection. So, and then they have to, after
these kind of collection, then they are culturing the bacteria on different
selective media, some Chrome Ager to see which bacteria are ESBL or CPO and
things like that. So based on that, probably only 10% of them are MDR or less
than that are MDR. And then depending on the physician decision, senior
consultant decision, whether any particular bug need to go further for whole
genome sequence, has the value for whole genome sequencing, because say
acinobacter, they don't think it's intrinsically resistant to CRE, then they
have no interest to find, to confirm that they have an intrinsic resistance. So
then probably people go for the gram negative, say CID, because they want to see
whether there's like a novel genetic elements, or they basically most of the
physician are not interested in ST, unless like it belongs to the SLI or ST
1193, they have no interest at all, because it doesn't communicate clinically in
terms of treatment. So I think practically it's feasible, but sometimes it
doesn't fit into their clinical workflow, because usually, I don't know, like in
the public health lab, in a reference lab, maybe they have a different workflow,
but in a clinical hospital setting, they have a set of protocol, like SOP in
making clinical decision, how these WGS technique, whole genome sequencing, AMR
genomics can fit into the decision making process. I think that's more
important. And I think at the moment, I don't see, I mean, I think the
physician, they don't see how, because, and also it's very, the operation cost
is really high comparing to a simple ComAGA test, because you're like
considering, like you have 6,000 cultures, I mean, on a weekly or even monthly
basis, we are not able just to use WGS to replace the conventional techniques
for screening. And also given the sample came from different kinds of specimen,
not, I don't think all the specimen are, you can do WGS on different kinds of
specimens, yeah. Maybe one day we'll just sequel, who's to know? Yeah, yeah, I
know. I mean, there are some pipeline that you can sequence two culture and then
to see like whether there's a mix, and even like I have seen papers in my co-
group, genomics saying that you can like, even as there's a mixed culture on the
petri dish, you can sequence all the material and based on the risk to assign,
but yeah. Yeah, I don't know. Maybe, yeah, I think it's feasible, but how we can
operate that. And not all the technician, like you can hire a high number of
technician who can do the conventional microbiology techniques, but you won't be
able to find that many technician who can do the WGS or the hospital is not
willing to. And also the other aspect is the accreditation process. I don't
think that at the moment, there's any licensing for the clinical staff purely
based on library preparation for whole genome sequencing. At least I'm not aware
of like, even in a cap in United States or in UK, I don't think there's a
licensing. So then many of the technician, if they need to require a license to
work in a lab, but if you only have the techniques to do molecular, you are not
able to get into the operation of the lab. I think that's another issue that we
have to work with the senior level. So yeah, both way, you have to standardize
protocol, then you get training and then people, technician get licensed to do
this kind of work then they can work in a clinical lab. Otherwise it's hard.
Grant, I'm gonna let Finley come in next. So it sounds like what the real
challenge is, is how do we use public health and infection control who are quite
keen on genomics, especially like break tracing, et cetera, as our Trojan horse
into medical microbiology, diagnostics, infectious disease treatment side. And I
don't know if anyone has any kind of comments or ideas on that. I do. There was
one hospital in the UK, they were being fined, they get fined if they have too
many C. diff cases. And so what they were doing was they were sequencing the C.
diff cases to prove that the C. diff didn't come from the hospital, but it was
acquired in the community to avoid getting a fine. So it is useful in that
respect. So Anthony Underwood. Yeah, I mean, I suppose in the public health
arena, we can afford to make some mistakes, not big ones, but it's not as
critical as treating a patient. So I think we can use the public health arena to
do the kind of surveillance kind of thing, look at the concordance, figure out
what's going on so we can get things better and learn our lessons in that way. I
guess the other thing, big thing for me, then something I've found during the
project I've been involved with is actually it's relatively easy running the
bioinformatics pipelines to process your genomic data to get a result, albeit in
a fairly sort of archaic format or hard to read format. The much, the bigger
challenge is going from that result then to what does it mean? And I think
that's where kind of that kind of intelligence encoded in a way which is not
just in someone's head and an expert's head, which is so often the case is where
the real challenge comes, particularly where there are nuances and you can have
particular beta-lactamases, which different alleles of the same gene can
actually be,  allow resistance to cephalosporins or carbapens, for example, and
trying to understand the nuance of that is very, very challenging. I think
that's kind of why part of this hackathon that I'm involved with, I'm trying to
bring together some of that sort of metadata into one pool so we can at least
look at that together. But there's a lot more to do. There's a huge amount to do
in terms of actually somehow distilling that intelligence and writing down a
series of rules, albeit may not be a simple decision tree, but maybe some kind
of rules which enable people to make that interpretation. I'm going to ask Lee
to come in here now because he's been very quiet. Definitely been very quiet,
but I'm not an AMR expert. I've just been listening and this has been a
fantastic discussion. There was one twist that I wanted to follow up on. It's
almost like that Jurassic Park quote, like, should we have been doing it? It
sounded like we might be venturing into the legal perspective with this. You
said that the hospital would want to sequence these genomes for liability
purposes. Are we going to feed into the legal system with this? Are we going
into a territory we don't really want to be in? That is a minefield and I don't
think we're going to solve that here. Particularly when you get into pathogens
like, I don't know, say HIV and whatever, could you go back and say he infected
me and blah, blah, blah. It could be quite a minefield. There was a study, I
forgot where, that was discussed at Applied Bioinformatics and Public Health
Microbiology. I think from the University of Maryland and they were studying
where all of their AMR, Klebsiella were coming from when they basically found
out it was coming from the pipes coming out of the sink. They couldn't get rid
of it. You guys remember that? I think they still have a whole sink lab, right?
University of Virginia or something? They have the PI from that project. I
forget her name, she set up a whole lab and that is now their project where they
have all these sink openings and they have plates and measure how far things
disperse. Yeah, biofilms in the drains and anytime you turn on the tap,
aerosolizes everywhere. Yeah, yeah. So could a hospital be sequencing all this
stuff and they can find out if it's their fault or someone else's fault? If it's
their fault, could they hide it? Not that I'm saying anyone's doing that right
now, but this is a whole minefield that we are not going to solve here. I agree.
Yeah, this is very sensitive. Yeah, very sensitive issue. There's a case in our
hospital that like we are a children's hospital, several babies are infected
with pseudomonas and then the physicians wanted to use WGS to confirm, to
investigate whether this is an outbreak. We did the WGS and then we find out
it's like a polyclonal. Our polyclonal is not like single clone pseudomonas.
They all belong to different sequence type of the pseudomonas. And then when
we're trying to report the data to the infection control, the meeting, actually
they didn't like it. Like we investigate this case at all. So in the end, all
the data is suppressed. We cannot carry on doing that. So it's blocked. So it's
very sensitive. Even internists, the hospital didn't really want the public
health senior, like if there's an outbreak in a hospital, they have to report it
to the ministry of public health, things like that. So it would affect the image
of the hospital in operation and also funding, whatever. I was going to say that
in the UK, we have a national health service. One hopes that there's a more open
and transparent and altruistic approach. But even there, now we have individual
trusts that kind of competing with each other. It's not in the interest of the
trust for everyone to know. It's your fault that you've got this organism that's
been traveling around your wards for a long period of time. In fact, when I gave
a public talk, well, a scientific talk about the asthma outbreak, asthma factor
outbreak that we had in Birmingham, one of the people in the audience said, you
know, hearing what you've said officer out around the back and shot them,
because they weren't doing their job properly. This is dereliction of duty. And
you sort of think, yeah, okay, you can you can play the blame game. And it does
become quite it could do. The only thing that I've noticed that when you're
dealing with clinical data and whole genomes and stuff, you can generally get
get it through. I mean, we did pilot, we published our ICU study published
asthma factor study and nobody in the in the trust headquarters said no, this is
reputationally damaging can't publish it. But when I've worked with people in
the animal health area, you're not you're allowed to know that this pig got this
brachyspira in the northeast of England. But the idea that you might actually
know which farm it was on or which town it was in, or any context at all that is
totally forbidden, because then the farmer would lose their livelihood. And it's
just so it could be more than human clinical medicine. I was just going to make
another point actually coming back to what Clement was saying, about the
technological side of things. It's interesting to contrast us trying to get
whole genome sequencing and even metagenome sequencing into the clinical arena,
whether with Illumina or now with Nanopore, with the advent of Moldytoff. And
Moldytoff basically rammed its way into the clinical microbiology laboratory,
just about the time I was leaving and wandering off into academic life. But it
actually it transformed things. And it became very much the that was the method
that people use to quickly identify organisms. And it's not digital. And so for
those of us who like our data to be digital, and like a tidy kind of universe,
it's like, what you're just looking at this kind of messy kind of wiggle
matching effectively. And you're you're trying actionable data. And they say,
yeah, but we can do that in minutes, you know, you can you give us a genome in
the same amount of time in the same price that's so cheap and whatever. So it's
been interesting that they're the competitor. And I wonder if there are lessons
to be learned. I mean, maybe if Nanopore becomes the next Amazon of genomics and
becomes a predatory monopolistic company, maybe they would be able to push it
themselves into the clinical lab. But otherwise, I don't know when it's gonna
happen. Anyway, maybe you edit that out. Well, the arena for a long time, but
but never made it. Yeah. Okay. Is there any final words? Or we'll go around?
Maybe we'll start with Finlay. Especially with the pandemic, the use of genomics
in surveillance, public health, infection control side of things really is like,
where it's been showing its use. I think the biggest some of the biggest
barriers is really getting to getting used more actively in the clinic is
getting it accepted in that more diagnostic mindset, that side of the clinical
labs, which I think is bound to be sometimes kind of can kind of homogenize from
the outside. It's like, oh, it's just the clinical labs, the medical
laboratories, some of the same people, they're just different hats on. And I
think, yeah, one of the big barriers is, like, integrating into that medical lab
tech kind of hierarchy and world and like, dealing with accreditation and all
that process that so people feel they can trust the data, and also communicating
results in a meaningful way that is actually actionable. And I think, yeah,
there's some great work kind of happening in those areas as well, coming out of
public health. Okay, Mark? I think an interesting point is that with whole
genome sequencing, we can look at where antimicrobial resistance is arising and
where the reservoirs are, and address some of these perennial issues. So the
medics will say, it's not us, it's all those people in veterinary practice, but
they're the ones that are actually making all the selective pressure to breed
this stuff up. And then the people in veterinary practice say, oh, it's not us,
it's the farmers who are misusing antibiotics. And it would be interesting to
have, and it is interesting to see, you know, how far are resistance genes
moving around from totally different contexts. And I mean, even looking at
hospitals, you know, hospitals put their sewage out into the outside world, that
sewage is full of antimicrobial resistance genes, and we don't know anything
about that. You know, does that matter? Does it not? There's lots of interesting
questions that can be addressed in a much broader context, rather than just
looking at... Thank you. Yeah, and I believe there's a study looking at sewage
and distance from hospital with vancomycin resistance found in bacteria, and it
was very close to related, unfortunately. Okay, Clement? Yeah, I do agree the
genome sequencing definitely helped to make that clinical decision. I think the
value is there, like I think that most of, I mean, although there might be some
limitation or maybe in terms of cause and also, but I think most physicians,
they like it. They like the technology and they like to have this kind of data
around, but probably who is making the call, like what I said, like usually they
decide which sample to be sequenced, not us. So, and I think, but also it's
important for the decision, they understand the concept and the technology and
also the power of that, and I think the pandemic helps in many ways, and they
like it, but of course, there's some technical and practical challenges. I think
there's still some challenge and limitation, yeah, but I do see the value. They
really like it, yeah, so I would say they like it, but sometimes given the
current framework and also the system, how to make better integration of this
technology into the daily workflow, probably that's something we need to
discuss, maybe even with the higher level, the national level as well, yeah.
Yeah, but I'm really optimistic for the future. I think when we look back to
where we've come over the last few years, we've made huge strides in actually
determining AMR, the phenotype of AMR based on genomics, I think it's looking
really rosy. I think there's still some big hurdles to cut, overcome, and there
was a paper by Ronan Doyle and colleagues, Katherine Harris, etc.  microbial
genomic massive discordance between bioinformatics prediction methods and so I
think there's a what we need to do better at is getting our benchmarking better
which is why it's great the hackathon has it has a benchmarking channel as part
of that and when we start to see better concordance and better agreement between
the different softwares and different databases then I think the adoption will
be much greater in the future. Okay thank you so much for joining me so that's
Clement, Mark, Finley and Anthony and Lee and we've had a great discussion today
and I think this is really an exciting area for the future and I hope that we
can speak again and we'll be that bit closer to getting genomics into the clinic
and being quite useful. 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 and 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.