Down To A ScienceLet’s talk
Menu

Down To A Science · Episode 01

Pathogen Genomics, AI, and Biosecurity

A conversation with David Aanensen

Hosted by Kevin G. Libuit · Presented by Galang AI

From beautiful software to better surveillance. How do we turn pathogen data into decisions—and put that capability in more hands?

Listen on your favorite platform:

The short version

At a glance

In the first episode of Down To A Science, Kevin G. Libuit talks with David Aanensen about turning pathogen genomes into useful public health information. From neuroscience and web design to Pathogenwatch, Microreact, and AMRwatch, David explains why intuitive software, representative data, and local ownership matter. They explore antimicrobial resistance, the shared foundations of surveillance and biosecurity, and how AI could coordinate trusted tools while keeping analysis transparent.

David Aanensen directs the Centre for Genomic Pathogen Surveillance (CGPS). This episode follows the connections between his career, the team’s software, and the infrastructure needed to make genomic information useful.

Ideas to take with you

Six key takeaways

  1. Capability matters as much as access.

    David defines democratization through the ability to generate, interpret, and act on data locally. A sequencer alone does not provide that capability.

    18:56 ↗

  2. Design can widen participation.

    Interactive trees, maps, and timelines let more people explore evidence and bring their own questions to a dataset.

    12:05 ↗

  3. A blank map is a sampling gap.

    AMRwatch makes uneven public genome coverage visible. Few available genomes do not establish that resistance or disease is absent.

    26:34 ↗

  4. Build surveillance capacity that lasts.

    David argues for connecting genomics to existing laboratories, sentinel sites, staff, and national systems, with funding for continued operation.

    33:28 ↗

  5. Sharing needs trust and local value.

    Countries and institutions need a say in access, interpretation, and the benefits created from their data.

    44:03 ↗

  6. AI orchestration is a practical starting point.

    The proposed use is to coordinate established analytical tools and data sources transparently. Predictive ambitions still depend on representative data and validation.

    54:28 ↗

Find your thread

Episode chapters

Approximate timestamps follow the supplied transcript. Each time opens that point on YouTube; edits to the published recording may shift the timing.

  1. 00:02 ↗

    Welcome to the inaugural episode

    Kevin introduces David and their connection through pathogen genomics.

  2. 00:48 ↗

    Neuroscience, sports cars, and the early web

    David’s route from biochemistry and neuroscience to a London interactive agency.

  3. 02:56 ↗

    Back to science: MLST and the web

    Brian Spratt’s lab, strain typing, and making genomic information accessible.

  4. 08:47 ↗

    Why scientific software should look good

    Pathogenwatch, visual design, and removing barriers to interpretation.

  5. 12:05 ↗

    Microreact and tools people can make their own

    Reusable visualizations, songs, sandwiches, and interactive scientific publishing.

  6. 18:56 ↗

    What democratizing genomics really means

    Sequencing access, interpretation, and the capability to act locally.

  7. 23:34 ↗

    What is antimicrobial resistance?

    Why failing antimicrobial treatments affect the foundations of medical care.

  8. 26:34 ↗

    AMRwatch: making the data gaps visible

    Public genome archives, quality control, geographic coverage, and community analytics.

  9. 33:28 ↗

    Build on the public health infrastructure

    Workforce, sustainability, WHO GLASS, national labs, and sentinel sites.

  10. 38:37 ↗

    Surveillance is biosecurity infrastructure

    Connecting public health and security priorities through shared surveillance capacity.

  11. 42:03 ↗

    PathGen, local tools, and data sovereignty

    Offline analysis and ownership of data, interpretation, and benefits.

  12. 46:14 ↗

    How trust makes data sharing possible

    A European MRSA project grows from country-level access to collective sharing.

  13. 50:48 ↗

    AI, drug discovery, and better data

    Laboratory validation, representative sampling, and the CASA collaboration.

  14. 54:28 ↗

    AI as an orchestrator of trusted tools

    Connecting Epicollect, Data-flo, Microreact, and Pathogenwatch into a workflow.

  15. 59:44 ↗

    Evolution, cancer, and model blind spots

    Why a model built on one population may fail in another; bringing in environmental data.

  16. 1:02:49 ↗

    Chytrid, amphibians, and a science-fiction tangent

    Fieldwork, microbial protection, and Alexander Titus’s Echoes of Tomorrow.

  17. 1:06:57 ↗

    Fungal genomics and the next scientific questions

    Reference genomes, complex fungal biology, AI research, and closing reflections.

The longer read

In-depth summary

00:48 ↗

A career shaped by science and visual communication

David describes studying biochemistry at the University of Salford, then neuroscience at the Institute of Psychiatry in London. A stint in a web agency brought him into a different world: building sites for Alfa Romeo and Fiat, collaborating with creative teams, and learning how presentation changes the way people engage with information.

Returning to science through Brian Spratt’s group at Imperial College London, he worked on web access to multilocus sequence typing (MLST). The role brought together his interest in evolution and his experience building digital interfaces. As sequencing and the web developed, the question became how to turn increasingly complex pathogen data into something a public health practitioner could use.

08:47 ↗

Beautiful, interactive tools change who can participate

Kevin recalls the impact of seeing Pathogenwatch demonstrated at an ASM meeting: a visual interface offered a different experience from assembling command-line tools and static figures. David makes a specific correction: the early demonstration did not include genome assembly. His broader point is that professional software engineering and careful visual design help scientists understand results quickly.

Microreact illustrates the team’s approach to reusable functions. Linking a tree, a map, a timeline, and metadata allows people to explore their own questions. David’s examples range from pathogen populations to song similarities and sandwich ingredients. The partnership with Microbial Genomics extends that idea to publishing: an interactive dataset lets a reader investigate beyond the figure and conclusions selected by the authors.

18:56 ↗

Democratization means owning the ability to act

Sequencing became more widely available during the COVID-19 pandemic, but interpretation remains a bottleneck. David frames democratization as ownership of the capability to detect, interpret, and respond. He also cautions that whole-genome sequencing is not always the best fit: targeted approaches can be more practical when a specific marker answers the surveillance question.

The conversation moves to antimicrobial resistance (AMR), where the loss of effective medicines threatens everyday infection treatment and care that depends on infection prevention. David emphasizes understanding which strains and resistance mechanisms circulate locally, then using that knowledge to guide surveillance, diagnostics, prevention, and the development of interventions.

26:34 ↗

AMRwatch makes missing information visible

David explains AMRwatch as a way to inspect what public genomic data can actually tell us. Sequence archives preserve research data, but they are not automatically representative epidemiological datasets. The workflow he describes brings public genomes through quality control and community analysis methods, then displays eligible data with time and location information.

The resulting maps reveal differences in coverage across places, years, and pathogens. The point is both practical and strategic: help people explore available evidence, identify where sampling is missing, and make a clearer case for investment. Counts in the conversation are approximate historical comparisons, not live totals or estimates of disease prevalence. The linked AMRwatch paper provides a dated, reproducible reference.

33:28 ↗

Public health and biosecurity share a foundation

David identifies reagents, workforce, laboratory infrastructure, and sustainability as essential parts of surveillance. He highlights WHO GLASS and the relationship between sentinel sites, national reference laboratories, and national reporting as a foundation on which genomic information can build.

Kevin asks how this connects to biosecurity. David argues that understanding what is circulating requires the same underlying surveillance capacity across public health and security agendas. His concern is fragmentation: separate initiatives can miss the opportunity to strengthen a shared system. AI can expand the questions people ask, but reliable, comparable input data still has to exist.

42:03 ↗

Trust, local analysis, and shared benefits

Asked about PathGen, David is cautious about concentrating data in a single system and emphasizes bringing analytics to the data. Local analysis gives institutions room to understand their data before deciding what to share. David describes an offline Microreact application as one approach, then broadens sovereignty to include ownership of data generation, technology, interpretation, and the resulting value. The discussion connects that principle to benefit sharing and the incentives surrounding international data exchange.

A European MRSA surveillance project provides the clearest example. David recalls countries initially asking to see only their own information. After working with those results, participants wanted to compare neighboring settings and agreed to wider sharing. In his account, co-development and a visible local benefit made collaboration possible. The separate Zika anecdote in this section needs qualification; see the editorial notes below.

50:48 ↗

AI can connect a workflow while preserving its methods

The conversation separates ambitious prediction and drug discovery from a nearer-term workflow opportunity. David points to César de la Fuente’s antimicrobial discovery work, while emphasizing the work required to validate a candidate experimentally. He also describes CASA and the need for structured sampling that improves both representation and local capacity.

For everyday surveillance, his proposed AI layer coordinates tools that already perform defined tasks: Epicollect gathers field data, Data-flo connects information sources, Microreact supports exploration, and Pathogenwatch processes genomic information. Kevin describes an LLM at the front end orchestrating deterministic components. The aim is to reduce manual transfers and time spent assembling an answer while keeping the underlying methods explainable. This is a discussion of a direction for development, not an announcement of a released end-to-end AI product.

59:44 ↗

Evolution, environmental context, and fungal blind spots

Cancer population biology prompts a comparison with evolving pathogen populations. David worries that models trained on a narrow geographic sample will not capture the mechanisms circulating elsewhere. He wants genomic information connected with other relevant data, including clinical and environmental context, rather than treating available sequences as a complete picture.

The final tangent moves from amphibian chytrid research and Kevin’s memories of salamander fieldwork to Echoes of Tomorrow, a science-fiction series. The imagined human outbreak belongs to the fiction. David then returns to human fungal pathogens, describing gaps in reference genomes and complex genome structures, including mobile elements called starships. The conversation closes with reflections on collaboration between scientific institutions and AI companies, and a jokingly bleak aside about AI’s future.

In their words

Transcript excerpts

Selected passages from the supplied transcript, with punctuation and capitalization lightly normalized. Timestamps identify the speaker’s turn; quotations may begin within that turn.

On democratization

“It's really about who owns the capability to detect, interpret, and act on their own data.”

David Aanensen · 18:56 ↗

On sovereignty

“So for me sovereignty means ownership of the data generation, the technology, the interpretation and the value that comes from it.”

David Aanensen · 44:03 ↗

On co-development

“It's co-development, co-development from the beginning, having the consortium of the right people and the right government agencies come together to define what that looks like towards a project that then gets owned by everyone.”

David Aanensen · 50:30 ↗

On AI and established analysis

“There's already hardened, deterministic, best practice ways for analysis, but sometimes getting them coordinated and orchestrated and understanding the decision of when to kick things off is sometimes the bottleneck at the day to day laboratories.”

Kevin G. Libuit · 56:16 ↗

Keep exploring

Show notes & source links

Tools, organizations, people, and further reading connected to the conversation. Each entry explains the connection; additional context is labeled. Links are references, not endorsements.

Software, visualization & genomic data

Organizations, surveillance & policy

More research, institutions & companies

People, research & the final tangent

Accuracy & context

Editorial notes

  • Names and terminology. These notes use Pathogenwatch, Data-flo, Auspice, Richard Neher, Reid Harris, and Batrachochytrium dendrobatidis where the transcript contains phonetic spellings. Oxford identifies the BDI director mentioned at 59:44 as Trey Ideker.
  • Genome counts are snapshots. The spoken SARS-CoV-2 and AMR totals are approximate and vary during the conversation. The AMRwatch paper reports 620,700 eligible genomes as of 31 March 2025. That is a filtered archival dataset, not a live total or a measure of population disease burden.
  • The early Pathogenwatch demo. At 10:39, David corrects the recollection that genome assembly was part of the early demonstration. Current capabilities should be checked in the tool’s documentation.
  • The Zika anecdote. At 44:03, David describes a vaccine developed using Brazilian sequence data and sold back to Brazil. This account is not verified here. WHO’s Zika fact sheet states that no vaccine is available for prevention or treatment. The anecdote should not be quoted as an established example of a commercial vaccine sale.
  • Fiction and forecasts. The human chytrid outbreak in Echoes of Tomorrow is a fictional scenario. The closing comments about AI and human extinction are conversational speculation, not a factual forecast. The discussion of AI-company laboratories is the guest’s commentary, not independently established reporting on this page.

Quick answers

About the episode & the show

What is Down To A Science?
Down To A Science is a podcast hosted by Kevin G. Libuit featuring conversations with practitioners in science, technology, and business about their work, personal insights, and real-world impact. The show is presented by Galang AI.
Who is the guest on Episode 1?
Episode 1 features David Aanensen, director of the Centre for Genomic Pathogen Surveillance, in conversation with host Kevin G. Libuit.
What does Episode 1 cover?
Episode 1 covers pathogen genomics, antimicrobial resistance, data visualization, public health surveillance, biosecurity, data sovereignty, AI workflow orchestration, and fungal pathogens.
Which tools are discussed in Episode 1?
The conversation discusses Pathogenwatch, Microreact, Epicollect, Data-flo, AMRwatch, Nextstrain, Auspice, Augur, Bandage, SPAdes, pangolin, and BLAST, alongside sequence archives and typing methods.
Where can I watch or listen to Down To A Science?
Down To A Science is available on YouTube, Spotify, Apple Podcasts, and Amazon Music. Episode 1 has direct listening links on this page, and the Show Notes index collects the episode guides.

Browse all show notes →

Privacy preferences

Change your choice at any time. Essential only keeps videos click-to-play. Videos you have already opened stay available until you leave or reload the page.

EssentialAlways on

Remembers your privacy choice so we don’t ask on every page.

No analytics or advertising tools are currently added by this site.

Privacy policy