Technological Advancements in public health bioinformatics
Summary
TLDRIn this episode of the Bioinformatics Lab podcast, Kevin Linwood and Andrew Page discuss the evolution of technology adoption in pathogen genomics. They explore the journey from software packages to cloud-based solutions, emphasizing the impact on interoperability and reproducibility. They reflect on the challenges of software installation, the rise of package managers like BioConda, and the game-changing introduction of workflow managers. The conversation also touches on the importance of open-source tools for public health and the potential of machine learning and AI to revolutionize the field.
Takeaways
- 🌐 The narrative of tech adoption in pathogen genomics has evolved from software packages to cloud-based solutions, impacting the field significantly.
- 🛠️ Early challenges included lengthy software installation processes and managing dependencies, which have improved with advancements in package managers.
- 📦 The introduction of Debian Med was a pivotal step, providing a sustainable and maintainable way to handle software packages in bioinformatics.
- 🔧 The significance of easy software installation was highlighted as a key factor in adoption, with tools like Homebrew and Bioconda streamlining the process.
- 🐍 Language shifts from Perl to Python have influenced the field, with Python emerging as a preferred language for its mathematical capabilities and community support.
- 🔄 Workflow managers like Make, Snakemake, and Nextflow have standardized the way complex analyses are conducted, enhancing interoperability and collaboration.
- 🌐 The adoption of cloud computing has been a game-changer, offering flexible, scalable resources that can be spun up quickly, without the need for extensive procurement processes.
- 🔒 Security and legislative restrictions have influenced cloud adoption, with some countries limiting data to specific regions, affecting the scalability of bioinformatics workflows.
- 🧬 The discussion highlighted the importance of open bioinformatics ecosystems, emphasizing the need for open-source tools and databases to ensure long-term support and accessibility.
- 🤖 The future of pathogen genomics is anticipated to involve significant adoption of AI and machine learning, which may change the field in ways that are currently hard to predict.
Q & A
What is the main topic discussed in the podcast?
-The main topic discussed in the podcast is the narrative of technology adoption in pathogen genomics, including software packages, containers, workflow languages, and the impact of cloud-based technologies on the field.
What does the podcast highlight about the early days of software installation in bioinformatics?
-The podcast highlights that in the early days, installing software could take a significant amount of time and effort, sometimes requiring a dedicated person to manage the process due to the complexity of dependencies and compatibility issues.
What role did package managers play in the evolution of tech adoption in bioinformatics?
-Package managers played a crucial role by simplifying the installation process of software and their dependencies, making it easier for users to access and use bioinformatics tools.
How did the introduction of workflow managers change the field of bioinformatics?
-Workflow managers allowed for the standardization and streamlining of bioinformatics processes, enabling users to string together different tools in a cohesive way, which improved interoperability and collaborative development.
What is the significance of containerization in bioinformatics?
-Containerization, through technologies like Docker and Singularity, has made it easier to manage software dependencies and ensured portability of tools across different computing environments, thus enhancing reproducibility and ease of use.
What is the impact of cloud computing on the field of pathogen genomics as discussed in the podcast?
-Cloud computing has allowed for faster access to computational resources, reducing the need for long procurement processes and enabling researchers to scale up their analyses quickly and efficiently.
Why is the adoption of open-source tools and platforms emphasized in the podcast?
-The podcast emphasizes the adoption of open-source tools and platforms because they promote accessibility, collaborative development, and long-term sustainability, which are crucial for public health and research applications.
How does the podcast view the future of AI and machine learning in pathogen genomics?
-The podcast views the future of AI and machine learning in pathogen genomics as transformative, with the potential to change the field in ways that are not yet fully understood, but will likely lead to increased efficiency and new analytical capabilities.
What challenges are associated with cloud adoption mentioned in the podcast?
-The podcast mentions challenges such as data sovereignty issues, where some countries have legislation that restricts the use of cloud services to within their borders, and the sensitivity of clinical data that cannot be moved across certain boundaries.
What is the significance of graphical user interfaces (GUIs) in making bioinformatics more accessible?
-Graphical user interfaces (GUIs) are significant because they abstract the technical complexities of bioinformatics, allowing users to focus on analysis and interpretation rather than command-line operations, thus making the field more accessible to a broader range of professionals.
How does the podcast reflect on the importance of open bioinformatics ecosystems?
-The podcast reflects on the importance of open bioinformatics ecosystems by discussing how they facilitate the distribution of tools, enable collaboration, and ensure that resources developed with public funding are available for public health laboratories worldwide.
Outlines
🌐 Evolution of Tech Adoption in Pathogen Genomics
Kevin Linwood and Andrew Page discuss the historical narrative of technology adoption in pathogen genomics, starting from software packages to cloud-based solutions. They reflect on the evolution from manual software installation to package managers like Debian and the impact of containerization and workflow languages on interoperability and reproducibility in the field. The conversation emphasizes the importance of making software easy to install to encourage usage and the challenges faced in the past with dependency management and software installation.
🛠️ Transition to Workflow Managers and Containerization
The discussion shifts to the advent of workflow managers, starting with make files and evolving to more sophisticated tools like Snakemake and Nextflow. The speakers highlight the pivotal role of workflow managers in standardizing analysis processes and enabling the seamless integration of tools. They also touch upon the transition from traditional HPC clusters to containerization with Docker and Singularity, which解决了 dependencies and portability issues, and the initial resistance and subsequent acceptance of these technologies in the field.
🌩️ The Impact of Cloud Computing on Bioinformatics
Kevin and Andrew explore the transformative effect of cloud computing on bioinformatics, discussing how it has simplified the procurement and management of computational resources. They recount personal experiences with cloud adoption, the ease of scaling up resources, and the challenges faced due to legislative restrictions in some countries. The conversation also covers the importance of open-source tools and the risks associated with closed-source solutions, emphasizing the need for open access to maintain and extend critical tools for public health.
🌱 Tech Adoption in Wet Lab Sequencing and Data Sharing
The speakers delve into the adoption of different sequencing platforms, from Illumina to Oxford Nanopore, and the implications for data generation and benchmarking. They express concern over the lack of inclusion of certain technologies in benchmarks and the potential biases this may introduce. The discussion also addresses data sharing practices, the challenges of cross-border data transfers due to security and privacy concerns, and the need for standardized outputs to facilitate data integration and analysis.
🌟 The Future of AI and Machine Learning in Pathogenomics
Looking ahead, Kevin and Andrew anticipate the significant impact of AI and machine learning on pathogenomics. They discuss the initial forays into AI with tools like Pangolin and the preference for interpretable methods over black-box models. The conversation speculates on how AI could change the field, potentially automating routine tasks and enhancing data analysis, while also raising questions about trust, interpretability, and the need for policy frameworks to guide the adoption of these advanced technologies.
🔚 Wrapping Up the Discussion on Tech Adoption
In the final part of the conversation, the hosts summarize their discussion on tech adoption, highlighting the dynamic nature of the field and the continuous evolution of tools and platforms. They express optimism for the future, with the anticipation of less manual work and more efficient data analysis through AI, while acknowledging the ongoing challenges in understanding and trusting AI-driven decision-making processes. The episode concludes with a commitment to continue exploring these topics in future discussions.
Mindmap
Keywords
💡Bioinformatics
💡Tech Adoption
💡Pathogen Genomics
💡Package Managers
💡Containers
💡Workflow Managers
💡Cloud Computing
💡Interoperability
💡Reproducibility
💡Public Health
Highlights
Narrative of tech adoption in pathogen genomics from software packages to cloud-based services.
Impact of tech adoption on interoperability and reproducibility in the field.
Early challenges in software installation and dependency management.
Evolution from manual software installation to package managers like Debian.
The importance of easy software installation for user adoption.
The rise of Homebrew and the shift to bioconda for easier software management.
Mamba's role in expediting the installation process in bioinformatics.
Adoption of workflow managers like make, snake make, and nextflow.
The transition from single tool integration to comprehensive workflow management.
The significance of Galaxy for accessible bioinformatics workflows.
The move towards cloud-based solutions and their impact on resource accessibility.
The critical role of open-source in public health and pathogen genomics.
Challenges and benefits of adopting different sequencing platforms.
The influence of cloud computing on the scalability and flexibility of bioinformatics.
The future of tech adoption with a focus on machine learning and AI in pathogenomics.
The potential of AI to transform bioinformatics workflows and decision-making.
The necessity for policy and trust in AI-driven bioinformatics.
Transcripts
all right welcome I'm Kevin Linwood
joined by Andrew page we're from Fijian
and this is the bioinformatics Lab
podcast
today we're going to be talking about
sort of narrative of tech adoption over
the years in pathogen genomics uh from
you know software packages to Containers
workflow languages uh and now at this
point of cloud-based guise and the
different ways in which that's impacting
our field
and are you having much more time in in
the field than I have so you have maybe
a longer time Horizon I say that with
absolute respect for your wisdom and
experience in the field
um but you have a broader perspective of
what this has looked like because again
whenever I talk about the sort of tech
adoption story I kind of started package
managers but of course there's beyond
that you know before uh but all this
conversation of how do we make this how
do we mature this as a field for
interoperability reproducibility and all
the likes so when you think about that
The Narrative of PEC adoption
surely it starts before package managers
oh yeah like I mean I remember the dark
days where you'd have to hire someone
you know just to sit there installing
software because it would be it would
take a week to install one piece of
software you know please be editing it
and trying to find 200 different you
know dependencies for a compiler that
you know was last released five years
ago and so we've come up we've come a
very long way
um even then you know like languages and
everything like I started off in I guess
Ruby and then Pearl
um for about Maddox and you know
languages change all the time and no one
uses Brill these days unfortunately well
me uh but anyway you know and then
python so you know things change over
time and uh we do get better and
actually I genuinely I think that um
Pearl
isn't as good for mathematics as python
you know I've after maybe about 10 years
I've come to that conclusion of years in
it but then I see like my son is using
rust and he's like oh this is amazing
you know like and uh it's much faster
and it probably is actually but I'm not
gonna give a python you know
um for some new fad language it's only
been around a couple years uh anyway so
I digress like it you know
so if you just take the most basic
installing software and package managers
like
the big thing a few years ago was just
you know with Debbie and uh yeah and uh
bunty's Debian packages so can you get
into can you get something in a Debian
package you know because once it's in
there then you're sort of for life you
know
um and Debian Med was the big one for
our fields which is um for biomedical
resources go in there
and yeah
phenomenally difficult if you start
packaging for Debian Med they will
assign you like a mentor if someone who
can guide you through the process over a
few months wow because it's quite you
know they want these things to to work
universally and you know be working for
a long time so a few months to guide you
through the process of building your
Debian packages you know which would
then be sustainable in the long term and
uh kind of easy to maintain and
gosh so that was probably first gun
during 2015 and because I just happened
to work with a guy who's one of these
maintainers and
that the advantage of that is when you
go to command lending and just have up
to get install blah like Rory or
something like that and that's that
lowers the bar so much that it makes it
trivial and if you make your software
easy to install then of course people
will use it and that's that's a key
magic trick you know to uh to getting
people to use your software is to make
it trivial to install and then you know
you go one step further and then we had
a Homebrew do you remember that yeah I
do that's that's about when I entered
the the chat because yeah which is uh
which is a little bit you know another
system they made a bit easier to go and
install software uh but it kind of fell
off the rails
um they became unwieldy condo came along
bio conda and that's been kind of the
default now for people to use
um because it just works and you know
everyone has kind of gathered around
that it's quite easy to install stuff
dependencies are you know are reasonably
okay at the moment and you can't get
into obviously a bit of a nightmare
sometimes
um and then Mamba has really helped the
installation process you know rather
than waiting an hour if tensor won't be
software because things got so complex
with resolving dependencies you know it
now works it's pretty quick so you know
we've come a long way
um we've also come along with workflow
managers what was the first workflow
manager used
make I I think I would consider make the
first kind of workflow manager I was
writing make files where I was like
writing shell scripts of you know just
executables and then I would compile
them to not compile them I guess but I
would kind of curate the workflow into a
make file and then I would Define
endpoints at the beginning and then it
would be you know that's how the
workflow worked it was just recipe style
it would Define the endpoint to find the
process is very much the same components
I mean and then that's where I started
learning about snake make which was
really make in Python
and then where I really started running
in workflow managers was uh use of next
flow and I think it was actually Aaron
young again on the staff b side who kind
of introduced us to um workflow managers
and specifically uh maybe was
Kelsey Floric actually I don't remember
exactly but
they introduced us to the concepts of
workflow managers so we kind of went
from the conversation of how do I get a
single tool
to work properly on my on my machine
ensure that the dependencies are
consistent throughout the different
environments but then workflow managers
changed it in that we were able to
string them all together in a really
cohesive way sorry one second
um but
that became kind of a really pivotal
Point again I'm speaking on the staff b
side but this is something that I think
is echoed throughout the field of it's
so much more standardized and how we're
doing these these these analysis you
don't have to write your own python
logic to string data from you know your
aligner to your variant caller to
whatever it is Downstream for
characterization
rather there was a consistent language
there somebody outside of it of of your
laboratory can look at it and understand
okay this is an excellent workflow I
understand that the architecture of this
this entire repository and moreover than
that I could take the modules from that
and build it into my workflow in a
pretty seamless way so that that was a
huge jump in terms of interoperability
collaborative developments that we saw
in staff B uh and we never really looked
back from that that became just a status
quo
at first one I used was um gold
vertebral resequencing uh code base it
didn't even have a name like VR copace
was the name and it was built for the
Human Genome Project not human that
hasn't Gina a Human Genome Project
um in the Second Street so this was this
was before people were doing things at
scale and genomics so you know you're
talking more than 10 years ago and said
I love you know next word didn't exist a
lot of these other things did not exist
yeah and uh so they had to be built you
know and uh sang as she was uh at that
time you know would be producing maybe
you know 20 30 of the world sequencing
from one place so you know they had a
scale different people didn't have and
so that's what we adopted and uh we were
done adopting it for pathogens you know
which is quite different to human
because you have to do a lot of things
uh you've got a lot of small little
things and that's very different to
human which is you've got a few big
things that you want to do things on
um so it's like you know it's a flipped
problem and of course that breaks
everything and uh yeah so anyway that
was the first one I ever worked on a
highly complex
um changing anything was very difficult
it was all in code you know I was on
GitHub whatever go to his own code and
it was all in Pearl as well and you know
everything was monitored by say writing
little files to disk you know and
checking them around whatever
so it worked but you know the growing
pains came in and so then the next
Generation came out
um
which again was a super lovely beautiful
pearl
um but you know again it was it was more
cloud computing focused and it did
things very well but but then you know I
had seen Galaxy a few times like oh this
is actually pretty cool so when I went
to Department that was first thing I
brought in was Galaxy because you know
you got a web server that anyone can use
it can run on us on a cluster or on
cloud and it's got all the tools you
know kind of LinkedIn and built in
because people you know again wrap-up
tools simple XML files wrap the tool up
and make it work and you know can
broaden wakanda or whatever and then I'd
like a lot of people to do complex
workflows a lot of people who don't know
the command line moving them away from
you know the this kind of black screen
and a blinking cursor and they don't
know what to do you know it's uh to oh
yeah I can just click here I can search
for
a liner and then suddenly you can learn
your data and you know everything Flows
In from the sequencers and then it can
self-service uh to do their thing and I
think I really do think uh gooey's you
know their worker weight in gold you can
see the amount of stuff that people can
do so even in Excel you know you've got
power users of excel it can do
phenomenal things if you just give them
something straightforward easy to use
good usability they can work wonders
when it comes to the command line
obviously that's very powerful for doing
things at scale and speed and for
joining things together but you know you
have a very high barrier for entry there
for for an ordinary person even even
someone who works in mathematics it can
be a high priority entry because you
might often have to read through lots of
documentation to figure out how exactly
does this fit in what exact format does
this come out in he shouldn't have to
sell that over and over again you know
it's nice when you can just kind of have
them linked together you know someone
sells it once maybe defines the outputs
say um for next flow or for for Galaxy
and then it's there and it's wherever
whatever more Works
um
yeah anyway uh
workflow managers are fantastic as are
things like self-independencies wakanda
I just love that I mean Jesus Christ
like the the pain the absolute pain you
used to have to go through to install
software and uh you know if you use
clusters like the kind of old school
HPC you know like basically physics
clusters from back in the day you know
massive you know machines in a big room
with blinking lights
actually installing stuff on those is a
right pain in the ass right because one
base operating system and you you kind
of install stuff and load and stuff you
know like the the operating system we
add that particular you know and
whatever is installed probably you know
five years ago that that's the version
you're stuck with that's the version of
compilers are stuck with and you can't
change anything you know so actually
having the ability to run Docker and
Singularity and whatever is is amazing
and that I remember the arguments people
had you know about allowing Docker to
run on a cluster like that no that's
security risk you know we can't be
having that Singularity did help as a
little bit with that but
um yeah it's uh we've come a long way
and now we're in the cloud like I mean
Jesus that's amazing yeah you can just
spin something up in a few minutes you
don't have to wait two years to go to a
procurement process and buy stuff you
can just bang There You Go have
resources run stuff and you have full
control over it then you can tear it
down and those are that's often the the
kind of three
in terms of tech adoption those are like
the three turning points I always have
in my mind it first is sort of maybe I'd
put a slash in package managers and
containers where it was solving the
problem of dependencies and install and
portability of their single Tools in
itself and that's solving a lot of the
problems that you described and then so
now we can all if I have a tool if I
have an assembler I can run it in a
really reproducible way in my
environment and you can run in the same
way in your environment then the sort of
next big Tech adoption was the workflow
managers not only do I have the same
tools I can stream them in a way that
makes it plug and play and as you
mentioned too these workflow managers
not only help to standardize that but
they also standardize the the running of
them I can run it on an HPC I can run it
on a local VM or I can run it in the
cloud and all the workflow managers now
have built-in capabilities
for that scalability and so that you
briefly mentioned it is that that third
layer is the the GUI the the web
applications that kind of wrap all these
things together you know and my first
experience with that was uh using Galaxy
actually through genome tracker they
adopted galaxies of platform and I
realized oh my goodness I don't have to
teach CLI any longer rather I could
teach them to click a couple buttons I
can show them the workflows and I can
focus on the results and the analysis
and interpretation rather than you know
your Linux environment your library
directory structure this CD means you're
changing folders or something like that
you know and and I think I'm stealing
this from I think it was uh Peter
um Van Houston in in Cape Town where
he's he sent a line and I don't think he
meant to be you know codify this but I
thought it was a really powerful line
and in public health
uh they should be doing less
bioinformatics and more public health
and I thought oh that's kind of a nice
little line it's like yeah we should
really be working to abstract the sort
of technical nuances of bioinformatics
and really instead of approach provide
them with tools that allow them to do
Public Health
but I think gooey's the graphic user
interface is behind that is is really
what what allows us to kind of Traverse
that problem where you don't have to be
card carrying by informatician to do
bioinformatics rather you can use
bioinformatics as a tool to inform your
public health that you're implementing
there and you know obviously we've seen
that I've mentioned Galaxy Tara has been
an incredible resource where you've seen
that kind of come to play
um and then you see again I'm still in
these terms from different people Ali
black and her 10 recommendations for
pathogenomics she coined this term of an
open bioinformatics ecosystem that's
built off all these Technologies we've
talked about containerized algorithms
that are written in the standardized
workflows that are made accessible
through these GUI portals and it's like
oh okay once you find that mix we've
seen that uh that model allow us to
distribute bioinformatics tools to
Laboratories across the world we've seen
that happen in so many different ways
you know we recently put out a
publication and where Tara from the
broad Institute has fit that model
though open bioinformatics ecosystem in
that it's got a really well maintained
graphic user interface Cloud backend
with gcp it's got the doc store
repository which is you know coming out
of the ga4gh universe there
um and then it's uh
because it's standardized workflows you
can also have the standardized outputs
where that you can then transfer these
outputs into different systems be it
transferred to you know SRA and ncbi for
for distribution and international
accessibility you can also transfer
these results into maybe more secure
environments where you might be
combining things with sensitive metadata
for genomic Epi investigation so
watching the tech adoption happen it's
gotten to us a point where now it's just
these resources that uh we've all been
coiling with and trying to make sure
that work on a machine now are in the
hands of laboratorians so that they can
use these things generate the results
and make sense of the data in real time
and form what's happening on the you
know either infectious disease side be
it public health clinical food safety or
otherwise
and you can see what happens where you
know when things go wrong when things
are not open like by numerics
um
the company has decided okay there's not
enough money to be made to maintain this
you know for the public health world so
we're just gonna shut it down and that's
it and that's a closed Source locked
away highly critical tool for public
health and it's it's going and had that
been open it would have been very
different because then obviously you
know the community could take it up and
you know keep maintaining it and extend
it and and whatnot and keep it alive but
if a commercial company holds the rights
to that you know the source code and how
it works and all the all the
infrastructure behind it then it's a
it's a problem and so we really do need
to have open mathematics open tools open
databases open everything
absolutely and that's definitely been
part of the ethos in you know how we
work professionally with Public Health
Labs is that we feel as if it's funded
and supported by public health and it's
applicable to a single Public Health lab
it's very likely going to be something
that other public health Laboratories
could also find utility in you know for
example we work with Laboratories in
Mozambique uh helping develop you know
assays for HIV sequencing and Analysis
these same resources that we're working
with aphl global Health to develop and
innovate upon in in Mozambique we're
watching being proliferated to
Laboratories in the U.S who have the
same interests and again getting back to
that ethos of we're going to develop
this tool publicly funded it needs to be
open source and open accessible because
it doesn't need to be closed off that's
not necessarily the business model
that's conducive to ongoing support
long-term of uh Public Health laboratory
pathogen genomics here
absolutely and under Tech Adoption Fund
actually I was at a conference day the
other day and the what was it
science meets policy using Next
Generation sequencing tackle foodborne
threats at the European food standards
Authority I know it's really really
eye-opening because all the different
you know each country in Europe you know
taking a slightly different way of doing
things but you know obviously everyone
is doing public health and everyone is
doing food safety and you want to eat
the same in goals more or less and it's
very interesting to see how people are
approaching different problems within
the context of their own country some
you know are going very much on the uh
I suppose a short reads you know buying
stuff I do CGM plus T shared at CGM LST
results and others you know very much
more well you know we'll raise the data
or we won't release the data or whatnot
and so yeah very interesting to see how
people are approaching the same problem
yeah Tech adoption it's big yeah we've
only really talked about it on the sort
of dry lab side of things but also on
the wet lab side of things how are
people adopting different sequencers
different platforms from Illumina to ont
I think you have some perspective on
high on tour and adoption and things
like this too
so yeah I was a bit horrified to see um
some Benchmark sets and they had like oh
and T sorry they had iron torrent data
but not ont then I was like come on like
you know I understand the last time I
saw one was actually uh is wrapped in
plastic waiting for a disposal in an
underground car park
it's the first sequencing data I ever
generated was actually on an ion torrent
PGM if you're familiar with that yeah
yeah
they had the little kind of Xbox logos
or something on them as well yeah they
even had I think a slot for your iPhone
if you wanted to you know put some uh
tunes on while you're preparing those
libraries it was it was an iPod or
something something similar that is a
really old connector as well yes yeah
yeah yeah yeah yeah it was uh that was
my first uh time uh generating
sequencing data we had those ion torn
chips uh I I I'm diverting there but
that was my first tech adoption into it
was iron tour and it was interesting
because you know again they're not the I
don't think we're seeing anything that's
wildly controversial but you know known
about the data but every time I would
put the data out there people talk about
the
the air profile kind of associated with
what was being generated there so it is
interesting watching benchmarks uh
getting generated where where you have a
single technology with a known error
profile without maybe also adding some
context with either Illumina data or on
T data especially for benchmarks
yeah and was very eye-opening was that
some countries don't allow people to use
it loud for security reasons
um so it very much limits them and what
it can do in terms of mathematics and in
terms of scaling up as well they can't
just go we need more bang There you go
that's a huge point in talking about
tech adoption is cloud computing that
because that was a big conversation is
infrastructure development across the US
that we were always a part of in staff B
and people were doing pretty much
everything you could see in terms of
on-prem servers hpcs Cloud working with
academic hpcs and all the like but then
it just became so obvious that cloud was
was the solution for all the reasons it
is in every other industry
um so that that's been a big thing but
but it took a while like even me in
Virginia it was like two and a half
three years of discussions with RIT
before we were given our AWS accounts
but I think we're at a critical mass
where we've seen so many Laboratories
kind of break the barrier have the
conversations with RIT that those
conversations are shortening
dramatically and you're seeing wide
adoption across really the world I know
in Academia like um when I came into the
courtroom it was all
um traditional HPC you know big cluster
in uh in a server room on-prem and we he
moved over to openstack you know which
is a private Cloud which is a million
times better and more flexible so you
know that's where we start today and I'm
sure in the future you know it'll be
become public Cloud because it's very
easy to go from an openstack private
Cloud to a public Cloud
um
but what general pandemic is fantastic
was that uh everyone all the um covert
sequencing was being done and uploaded
to uh an academic Cloud called climb MRC
climb which is uh you know openstack
based on three different universities in
the UK and so instantly resources are
available you know like no one had to
sit around and say oh you know can we
rent some resources from here or borrow
some from here it's just like fine there
you go
um all available and then the public
health authorities
um like ukhsa we're using azure
um and you know everyone was just it was
just Cloud light loud you know in a
private public and there's no sitting
around you know waiting for to to buy
vasm at the storage was just like well
here's what we can right now and solve
these problems I know other countries uh
struggle because there's legislation
that says I don't know they can only um
if they were to use the cloud it can
only be within their country or within
one particular region so it kind of
makes things a bit more difficult
obviously Amazon Google and Azure uh
some Microsoft are doing a good job by
having data centers all around the world
but it is a problem and then when you
get done clinical stuff a lot of stuff
can't even leave like the hospital you
know something really sensitive data you
know you just can't cross that barrier
when out to Quantum we uh the building
had the
we shared with local hospital and
obviously research in hospital we had
two separate physical networks in one
building and one you know one kind of
server room as well you know two
separate physical networks because they
had to keep you know the clinical stuff
Toki separate from the research stuff
researchers seen as you know I've seen
more Lucy goosey and high risk than that
clinical stuff well I just fair enough
but anyway it hopefully with time people
will adopt a cloud a bit more
and before we end the last front or I
guess we talked about with the
historical Tech adoption now looking
forward I think you know you and I have
had a couple episodes on this but what
we're seeing is the big Tech adoption
that's really going to be changing the
field is you know machine learning and
Ai and how it's going to impact things
and we saw a little murmurs of it even
like with Pangolin there was the pangol
learn and it was like okay how are
people going to deal with this sort of
black box of decision making and it
wasn't taken to Kylie I mean it was
taken in its utility and practicality
and speed but it was definitely a huge
preference to the phylogenetic placement
because people could make sense of the
Usher uh approach versus machine
learning but it's going to be
interesting over the next couple years
where more of this adoption of machine
learning and AI in general there
um or more specifically I guess it is
going to be interesting to see how uh it
impacts our field I don't know if you
have a hot take in the last couple
minutes before we close out in Tech
adoption of AI and uh in pathogenomics
I honestly think it's going to change
things in ways we can't even comprehend
right now you know it'll just be there
it'll always be on it'll just be part of
what we do and I'm excited for the
future you know I'm excited to do less
work I'm I love now when I program that
uh it saves me so much time having to
type stuff out because you know like it
can you know pick up all the obvious
stuff that I I'm probably gonna do
yeah because we're already using it in
active writing active programming I'm
really interested to see how the big
tools kind of come in here because a lot
of the what we do is a lot of classic
categorization that seems well for the
taking for these kinds of Technologies
but I think it's going to be the policy
and Adoption of how do we
to what level of trust we put to this in
knowing that we can't really necessarily
just open up and understand how we came
to these decisions so
that's what keeps our jobs exciting
absolutely all right good episode we'll
uh continue on this conversation I'm
sure in future ones in the coming weeks
Voir Plus de Vidéos Connexes
Creepy or Cool? Exploring Personal AI - PI AI Blew Our Minds!
Reviewing the AI Battlefront | The Brainstorm EP 38
Andrew Ng - Why Data Engineering is Critical to Data-Centric AI
ひふみ目論見倶楽部【準備運動編】安宅和人さんと一緒に未来について考えました。
Code Interpreter.... but OPEN SOURCE? Open Interpreter's Mike Bird on OS Projects, Mindset + More
Silicon ANGLE the CUBE Executive Series: AI Infrastructure Silicon Valley
5.0 / 5 (0 votes)