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's 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. Once again, the Canadians have taken over the podcast and
will continue to talk about their public health response and genome epidemiology
of COVID-19. We cross back to their conversation now. So far, we've talked about
tools. We've talked about variants of concern, and no podcast would be complete,
especially one that Will and I are involved in, if we don't talk about metadata.
As we've already discussed, there's a lot of players that are contributing
information to our National Genomic Surveillance Database. There's a lot of
different metadata streams coming into that database. So how is the metadata
being harmonized for analysis? Will, I'm looking right at you. From my
experience, I was a biomathematician and a scientist in the public health lab
for a period of about eight years. So the challenges associated with data
sharing in the public health system is something that's highly familiar with,
but also sympathetic with, because it's not just people are not sharing data.
They're not sharing data often for the right reason, in terms of privacy and in
terms of logistics, right? To prepare the data for sharing often consumes a lot
of valuable time, especially during pandemic. So early on in the CanCoGen
effort, we divided the metadata into three different tiers. We sort of label the
low-risk, de-identified information, the minimum metadata that can be released
publicly and used by researchers right away. And then we have a middle tier that
consists of a bit more than just the minimum metadata that can be useful for
national surveillance and national tracking. So in other words, data that will
be shared between the province and the national lab. And then we have what we
call the local metadata, in other words, metadata that are identifiable that
each of our partners are encouraged to keep for themselves. The data, of course,
will later on be valuable for research as well, but we might not have time to
de-identify them or process them for research, but we nevertheless encourage
these data types to be kept in a standardized format and therefore can later on
be de-identified and used for research purpose. So we established these three
tiers of data grades and that formed the framework of the CanCoGen metadata
management plan. Little did we know that the minimum metadata part actually took
a few months to negotiate. We realized that what one province or one
jurisdiction defined as non-identifiable information or de-identified
information, namely, for example, age and gender and the date of sampling,
normally considers non-identifiable given that there's sufficient number of
people sharing the same, nevertheless become a point of discussion over a period
of a few months within Canada to sort out. And we even had to consult legal
experts to help us understand whether there's any legal barriers saying that
these data types indeed constitute identifiable information. And we end up
writing memos to the public house to help them understand that the global
consensus is that these types of information are indeed not considered
identifiable. And the release of such information, as we know by now, can help,
especially the date of sampling, can really help with global tracking of the
viruses and to understand the transmission of the virus. So that was sort of our
metadata organization sort of framework. And in terms of the metadata
harmonization, indeed it's very much Emma who's leading the work as part of the
CanCoGen in the PHAGE consortia. What does PHAGE stand for? So PHAGE stands for
Public Health Alliance for Genomic Epidemiology. You think a working group chair
would know that, Will, don't you? You might think that. Emma. Oh, right. I think
she, yeah. So she's trying to quiz me rather than Matt Norman. Yeah, yeah. It
was no fun to let her get away with that like that. So- Hey Matt, I'm just
asking the questions here. You guys are the ones in charge of that. Very much
early on, we have established a long list of metadata that are relevant to SARS-
CoV-2 sampling and have published that as part of the PHAGE consortium and of
course adopted for the CanCoGen use. So within that large data set is the
minimum metadata that I mentioned before, but also the overall standard consists
of additional data fields that we deem relevant. And this is very much done in
consultation with academic researchers, with public health workers and other
types of stakeholders that are involved in the pandemic response. And over time,
over the last year, we have had additional fields to keep track. For example,
the reason for doing sequencing, which can help with the interpretation. In
epidemiology, the reason for doing a sample collection or data collection and
the so-called denominator used in the calculation is it's often very key to the
analysis. So tracking information about why samples were collected and the scope
of sample collection and the inclusion, exclusion criteria and so on. Much
things that we over the time added into the standard to help improve the
granularity and improve the details for information and the standard have been
made publicly available. And indeed we work in the open access, open data
framework to ensure that these standards are widely available and anyone can use
it. And indeed anyone can contribute to them and help us improve these
standards. And also within the CanCoGen project, we have established a group
that programmers are helped to develop tools to help standardize the data
collection process. So in my group, we developed a tool called the Data
Harmonizer that can be used to capture the metadata in standardized format and
it has built-in validation to ensure that the data is in the format that
consists with the standard and also has export functions that allow the data to
be exported in a format that's compatible with public repositories and other
organizational needs such as within CanCoGen reporting to the national lab.
There's a specific format that's needed. Metadata Harmonizer essentially handles
the data conversion for you. And as I mentioned before, some of these data
cleaning efforts are time consuming. So these tools are designed to help
streamline the efforts for data standardization, harmonization. And also Emma
leads a curation team that help stakeholders to manually clean up the raw data
and do the conversions. Sometimes with the help of the Data Harmonizer and
sometimes indeed the data are so incompatible that they need to be manually
harmonized. And we have a designated team that help to achieve that so we can
have high quality metadata. So I just want to give a quick shout out to the
previous MicrobeMP Podcast episode where we discussed the phage metadata
standard. We'll link to that as well as all any of the tools that we've talked
about in this episode in the show notes. So that if you're interested in using
that standard, you'll be able to. So at this point, I want to get maybe a little
controversial. I want to address an elephant in the room. So Canada has been
sometimes criticized in the media and occasionally by its own scientists for the
slow pace of data sharing with public repositories. So I'm going to lob this
grenade over to the both of you to comment on that, particularly why that might
be the case and what are the steps that are being done to improve the situation.
So I don't know who wants to start. I would say it's not just the sharing of
data at all to the public databases, it's also sharing incomplete metadata,
particularly in terms of incomplete date metadata. Art Poon has done some
really, at Western, has done some really nice plots that demonstrate that among
the Western countries that have been doing a lot of sequencing, Canada has a
particularly low percentage of their genomes in GISAID that have complete date
metadata. And that varies hugely by province as well. So it's really, again,
it's a product. I mean, I'm an immigrant to Canada, so I can't badmouth it too
much because they could still deport me. So they don't have to give me
citizenship. I think it's largely a product of that kind of fragmented, having
all those individual provinces that all have their own legal systems, all have
their own data privacy rules. And then you have the federal government, which
has different relationships with each of those provinces, has stronger and
weaker ones, more contentious ones based on, you   which governments have put in
charge, for example. And so that kind of adds a whole other layer to that. And
there's a lot of provinces that are looking forward that maybe are interested in
being more independent than others from the federal government. So are keen to
avoid anything they see as a reach of federal government power and enforcing the
federal government power. So they're kind of looking to the future. Although
Will, I think has been at the sharp end of a lot of these discussions and issues
more than I have been. Yeah, so Will, I'm sure you have a few things to say
about this. I think that the key challenge indeed, as Finn mentioned, is that
there's different privacy laws in each province. And in addition to that, each
province has its own privacy officers and lawyers interpreting these laws. So to
derive a consensus that can be shared across Canada has proven to be challenged
time and times again. There are previous attempt to try to come up with
multilateral agreements for data sharing in public health. And all end up being
held up by difference of opinions across the country. And indeed, we are seeing
that again during the COVID-19 response. Some provinces are very quick to share
data. Other provinces are willing to share data, but has reservations such as
release of state information to its full extent. Because the potentially, well,
I'm actually not sure exactly who's making the calls, but the privacy officers
or the public medical health officers and so on may deem those data to be
identifiable and therefore refuse to release that publicly. And Canada as a
whole indeed is, what I would characterize as more risk averse when it comes to
releasing of information into the public. So contrast to countries like Denmark,
and some other countries that are quick to release these case related
information in a de-identified manner. Canada overall just has more, taking a
more cautious approach. But while that's the main problem, indeed elephant in
the room, some of the considerations that came up during the discussion
nevertheless, I think it's valid. For example, at the beginning of the
discussion, attentions were paid to the quality of the data being released. So a
lot of efforts, as Finn has already alluded to, went into the QC of Canadian
datas, that the hosting of Canadian datas to ensure that the data released of
high quality and that include both sequence data and metadata. So that also
contribute to the, legitimately contributed to the delay. I think there's no one
major reason that we overall are slow to respond, but many, many different
reasons. And we have addressed them one after another as part of the CanCo chain
coordinated effort. And I do think moving forward, we'll see the Canadian data
being released much more quickly. And hopefully as part of this COVID-19
response, we will come up with a better system for releasing of sequence data
and the minimum data for public, into the public repository for global pandemic
or global outbreak response. What's kind of interesting in Canada and the kind
of thing that we haven't talked about is the legal framework in the federal
government does actually give the federal government the power to compel
provinces and territories to share all their data. The Public Health Agency of
Canada Act, set up in 2004 after the first SARS pandemic, as well as Statistics
Act. So those powers exist, but the problem is so much of the function of Canada
is based on there being relatively good working relationships between different
provinces, different territories, as well as sovereign first nations bands
across Canada. That's a whole other aspect of kind of interesting complexity, by
being a settler nation that encompasses many other sovereign entities. There's a
whole other layer of kind of international diplomacy within Canada, essentially.
So that adds a lot of kind of complexity to that. And there's a lot of,
especially amongst some of those communities, a lot of reasonable hesitation and
worry about the federal government exerting power. There's kind of an
interesting dance going on between, you know, we're not going to compel these
provinces. We're not going to use these acts to compel the province to do this
if you share the data with us. We're not going to force you if you do it. So
there's an interesting, that kind of dance. But it does, like, there's a lot of
criticism about this. And there has, you know, some people have gone as far as,
you know, the international health regulations set up by the WHO after, you
know, after the subversive pandemic. Basically, you know, that all member
nations must be able to share, you know, detailed epidemiological information
with WHO. It's Canada, due to this provincial-federal gap, is actually in
violation. Some people have gone as far as making that argument. So it's a huge,
huge problem, but it's not one that's got really one easy, quick solution
because it requires all of these moving parts to kind of move in concert without
anyone getting trodden on in the way. At the end of the day, the public health
system is to serve the public and health officers at both federal and provincial
level, making calls on behalf of what they think are the best for the
population. But news medias and so on, time and times again, highlighted that
the Canadians do think that the data-sharing challenges that are seen in Canada
are detrimental to the pandemic response. So I think there's an opportunity
there to really assess indeed what are the public opinions when it comes to
sharing of some of these minimum de-identified information publicly and timely
for the combat of infectious diseases. And that would allow potentially freeing
the public health authorities hand a bit more because they know that they're
doing it in accordance to the population's desire and wish. I think that's a
critical missing piece and that's why everyone has been a bit more risk-averse,
not wanting to be the one accused of breaching privacy or breaching people's
trust. So to sort of look towards wrapping things up, looking at the overall
picture of SARS-CoV-2 sequencing in Canada, what are the things that you would
say are working well and what are the ongoing challenges? You talked a lot about
these things throughout the podcast today, but in summary, what would both of
you say are the things that are working well and what are the sticky wickets
that we're looking at? I mean, I would say one of the things that really is
working well is essentially every time we've kind of cross-cut that particular
problem of the federal provincial systems and all those things, by having things
like Cancun, having these different working groups and having just ad hoc
conference calls where all the people, even from, say, a province that doesn't
like sharing data and really doesn't like the federal government, but all the
people working in the health labs there are working hard on the same problem and
are eager to discuss things with their counterparts in other provinces. So
there's some really great kind of discussion and conversation going on about
that. And I think it's good. There seems to be a lot of discussion about kind of
looking to the future, about trying to build more long-lasting infrastructure
for this and really scaling up the use of genomic epidemiology, which as three
people who have all spent a lot of time trying to get genomic epidemiology
working, you know, more broadly across Canada for foodborne disease and so on is
a good thing. And, you know, individual hospitals, like how do we ramp up our
use of genomics? You know, this has been really important or really useful. So I
think that's one of the really good things. I agree. And I would say it sums up
to, this is a trust-building exercise and we are indeed through this process
know and highlighting some of the barriers and challenges. And I think we indeed
need to overcome them and build better trust within the Canadian public health
system. So I think that's what's going really well, the communication and the,
we all know that we spend disproportionate amount of time in meetings and so on,
but in a way those are necessary in order to, for a large group of people in a
large country to achieve consensus and to share expertise and share knowledge.
And I see that sort of really come across in the CanCoGen activities. What I
don't think working very well is we still have that gap between what the
practitioners are working on and what we are sort of, you know, working
tirelessly, trying to build pipelines, working tirelessly, trying to build tools
for sharing data and so on. But we still have that gap of understanding in terms
of what are the need to be put in place in order for data sharing to occur
across Canada. What are some of the guidance documents or what are some of the
regulations that need to be put in place? And as we mentioned earlier in this
conversation, indeed that Canada has to improve its overall framework for public
health data sharing in order to be more responsive in the future. And for that,
I don't think we've done enough work to understand that process. And within
CanCoGen, indeed that has to be the focus for our sort of year two of efforts,
trying to understand the legal, the governance, the ethical considerations,
trying to understand the public health, sorry, the public opinion on public
health data sharing better across Canada. And by improving the social science
aspect of data sharing, I think that's when we can really bring the technical
works that we have put in place.  into its full utility. Great, and so in
bringing this episode to a close, I just want to ask you one final question that
probably overlaps with the last question, but if you were advising a country
with a decentralized health system that was starting to build capacity for SARS-
CoV-2 sequencing, based on the lessons that you've learned, what advice would
you give? I mean, the first thing is centralize as much of the, I think,
analysis and data collection and curation as you can, just so there is at least,
not necessarily even in like there being one final repository, but there just
being some form of centralized data repository, some form of shared analytical
platform. And there's lots of different ways of doing that. You know, there's
things like IRIDA, you know, that we've talked about. There's, you know, SP3 for
tuberculosis. Like there's kind of a lot of different ways of trying to do that.
I would say the best way and the way that we've tried to deal with that
fragmented decentralization is really trying to focus on QC metrics and having a
robust set of QC metrics that are clearly communicated, but then also checked in
that centralized fashion. So the genome is being checked in a centralized way.
And then third, I think largely just try and do what the UK did, but try not to
have a Tory government. Like the great, you know, the genomic epidemiology side
of the UK, many people listening to this are involved in that effort, has been
excellent. It's been really world-leading, but the gap has been in the way that
that information is being used at a government level, at a political level to
actually implement policy, I would say. And that's where there has been major
issues. So yeah, trying not to have a conservative government tends to help.
Yeah, well, indeed, you look at the countries that responded well to COVID-19
versus countries that respond poorly, there is a high correlation between the
strong national leadership when it comes to, you know, disseminating expertise
and policies, right? And those are actually, ironically, the strong national
leadership comes from, more likely come from a liberal government than from a
conservative government, right? Usually you associate strong national
centralized leadership with conservative government, but it's indeed actually
the liberal government that provide much more leadership in this area. And so I
would echo Finn's comment, even though I don't know if that comment about the
Tory government is going to be, it's either going to be made the tagline of this
talk or will be cut out, so we'll see. I mean, if you have a government that's
already expressed its contempt towards starving people and allowing ongoing use
of food banks and unprotection of vulnerable people, such as the homeless, is it
surprising that on a policy level, when there's a pandemic, those same
vulnerable populations are going to be particularly left exposed? And I'm not
saying Canada's done a particularly great job at this overall either. You know,
there's been a huge issue of not protecting vulnerable populations and not
really factoring in SES. There is a lot of work trying to go into that, but one
of the problems we have in Canada is environmental. So we have an issue of, you
know, very cold winters. So when you have shelter systems that then have to
reduce their capacity to increase social distancing, you have to have people
making a choice between a vulnerable population with comorbidities, that's case
fatality rate around 10 plus percent for COVID-19, versus what's the case
fatality rate for freezing to death? So really, that is where, like, you know,
we're a German company always, we're bioinformaticians. Like, we're in the data,
but like, there is a point where that's actually directly impacting policy. And
that final link of the chain is something I think we should be trying to get
more involved in. And I think that's actually a key part of trying to build a
decentralized health system using genomics, is the domain experts actually have
to get far more involved in that political side, far more involved in that
implementation side. Because otherwise, A, we can get divorced from the entire
process. We can lose track of what we're actually doing and whether it's, like,
the impact of even our mistakes. Right. And we're gonna have made mistakes in
this process when we do all the policy work. Yeah, okay. And using current
technology, indeed, the strengths of web-based technology is that you can have
highly connected decentralized system that still communicate and functions well.
But the key there is to have, you know, a trust framework that enables data to
be shared in that decentralized system, rather than setting up silos in
decentralized system. So my advice, and it doesn't have to be top three things,
is that if you have decentralized healthcare system, you should have a well-
connected decentralized healthcare system, the one that functions through
sharing of data, sharing of information, and sharing of expertise. Okay, so I'm
gonna end things on those calls to action. I want to thank Will and Finn for
sharing all their expertise and their knowledge and their great banter here
today. I'd also like to thank Lee, Nabeel, and Andrew for inviting us on the
show, as well as all of our Kinkagen partners and all the hardworking frontline
workers everywhere who sacrificed their time, energy, and safety to keep us
safe. So if you want to get ahold of us to continue these discussions, you can
tap us on the shoulder at the microbenfi slack, or you can contact us by email,
where we'll be sure to pop all of the links to any of the tools that we
mentioned today in the show notes. And with that, thanks everyone, and stay
safe. Thank you all so much for listening to us at home. If you like this
podcast, please subscribe and like us on iTunes, Spotify, SoundCloud, or the
platform of your choice. And if you don't like this podcast, please don't do
anything. This podcast was recorded by the Microbial Bioinformatics Group and
edited by Nick Waters. The opinions expressed here are our own and do not
necessarily reflect the views of CDC or the Quadrant Institute.