And now, a mega-thread: If you've ever wondered what connects all our work at @viralemergence, our new paper in @NatureMicrobiol ties it all together. No, really. It's all one thing. Want to step through the Verena Cinematic Universe together?
Our team uses big data, statistics, and machine learning to understand "the science of the host-virus network", a broad methodological problem that includes a number of smaller, more applied problems. nature.com/articles/s4156…
If you want to study the host-virus network, you need data. But as @roryjgibb & co. showed, existing datasets are full of taxonomic inconsistencies and conflicts. We needed a synthesis. academic.oup.com/bioscience/art…
That's why we built Global Virome in One Network (VIRION), the most comprehensive (and clean) atlas of the vertebrate virome ever developed. biorxiv.org/content/10.110…
Our work doesn't stop there; @taddallas & co. built a new interface to the Ecological Database of the World's Insect Pathogens, one of the few resources available for the insect-pathogen network. ecoevorxiv.org/yd3x5/
What can we use those kinds of data to do?
(1) Predict the host-virus network!
@tpoi & co. propose a new method to recover the "missing" network (>90% of it), especially in the Amazon basin, which has been severely undersampled for wildlife viruses. arxiv.org/abs/2105.14973
@taddallas & co. have also recently shown that you can extend the host-virus network problem to predict the host-vector-virus network for mosquito-borne flaviviruses ecoevorxiv.org/xzmp8/
We've been able to use our host-virus network predictions + network embedding to create the most accurate artificial intelligence ever developed for zoonotic risk prediction. No, really. arxiv.org/abs/2105.14973
The backbone of that model was just published by @NardusMollentze with his own group, and I can't recommend this paper strongly enough. A masterpiece of predictive research. journals.plos.org/plosbiology/ar…
We've also written as a team about where tech like this might start to be useful for global health security, building on a workshop that brought together bench virologists, computational experts, and global health practitioners. royalsocietypublishing.org/doi/full/10.10…
In the future, it'll be important to go beyond zoonotic potential and also look at epidemic potential, including transmissibility... journals.plos.org/plosone/articl…
I should say - we owe so much on this to @bahanbug, who developed the foundational approach: identify what animals have some viruses of interest; use machine learning to predict which others might also have those viruses. journals.plos.org/plosntds/artic…
At the start of the pandemic, our team (led by @danjbecker and @Gfalbery) built the first multi-model comparison study to predict potential bat reservoirs of undiscovered betacoronaviruses. biorxiv.org/content/10.110…
We've spent the last two years proving that we can, actually, use artificial intelligence to optimize the search for new betacoronaviruses. Now we know that they work best when we actually use data on species "traits" - their ecology, evolution, immunology.
We also increasingly think that these predictions can be improved by using data that better represent reservoir competence for viruses (e.g., viral isolation instead of just PCR or serology) cell.com/trends/ecology…
In some of our collaborative work with the Forbes lab predicting rodent reservoirs of undiscovered orthohantaviruses in the Americas, we see that improvement when we model viral isolation vs. PCR: biorxiv.org/content/10.110…
We also think tools like this will increasingly help us anticipate pathogen "spillback" from humans into wildlife. This review by @annafagre & co. outlines the theory behind the approach, as well as the data we'll need... ecoevorxiv.org/sx6p8/
...and this new study by @bahanbug's group, which uses ACE2 sequences to predict potential wildlife hosts of SARS-CoV-2, is proof-of-concept that this could work! biorxiv.org/content/10.110…
4. Viral diversity: why do some animals have more viruses? Why do they have more *zoonotic* viruses?
Things get a bit complicated here. In some of my own work we've shown that we probably only really know about ~1% of the mammal virome (50k+ viruses). nature.com/articles/s4155…
@roryjgibb & co. showed that this creates a bit of a problem: where we look for viruses, we find them. And if we ask questions about viral diversity, the answers change over time. biorxiv.org/content/10.110…
@Gfalbery & co. showed that some key ecological hypotheses don't hold up after you correct for sampling bias. Sure, animals in cities have more known zoonotic viruses - because they have more known viruses, because we're looking more. biorxiv.org/content/10.110…
Results like these are an unexpected challenge to the "pace-of-life theory" - i.e., the idea that fast-lived species like rodents make immune investments and invade habitats in ways that predispose them to zoonotic risk. @Gfalbery and @danjbecker explain: cell.com/trends/parasit…
@Gfalbery@danjbecker We've also got some preliminary results that take this further, showing that domestication probably increases zoonotic viral richness, but involvement in the wildlife trade doesn't. github.com/viralemergence…
@Gfalbery@danjbecker 5. Which hosts share viruses with each other? @Gfalbery & co. showed that - like dozens of studies have shown at smaller scales - it's your proximity to other animals in geographic space and evolutionary time that determines how similar your viruses are nature.com/articles/s4146…
@Gfalbery@danjbecker@NatureMicrobiol Together, @Gfalbery and I have applied the same viral sharing model to predict how climate change-driven range shifts could completely reshape the global virome, creating hotspots of cross-species transmission in the places where we'll live in 2050 biorxiv.org/content/10.110…
@Gfalbery@danjbecker@NatureMicrobiol I'm really not kidding when I say "it's all one thing": predicting the next pandemic virus, tracing SARS-CoV-2 to its origin, projecting hotspots of climate-driven disease risk - it's all the science of the host-virus network. That's what we do. nature.com/articles/s4156…
• • •
Missing some Tweet in this thread? You can try to
force a refresh
One last thing before I drop off for the holidays… a bit of behind the scenes on our new paper
Before there was a thing called Verena, there was me and @Gfalbery, some long chats over beers, and 12% of an idea for something called VirusNet - a team that would stop duplicating efforts and pseudoreplicating the same analyses of the host-virus network, and go far together
It’s absolutely bewildering to think this is the first thing @viralemergence worked on and that it’s finally, finally out there.
As a companion to our new paper in @NatureMicrobiol we've opened up the Host-Virus Model Database, the @viralemergence team library of studies that try to predict the host-virus network. How does it work? 🧵
In our new paper, we define a taxonomy of six types of models that try to predict the host-virus network; in practice, they don't always look and feel like network questions (e.g., do some mammals have a higher richness of zoonotic viruses?) nature.com/articles/s4156…
We outline six big model "shapes": predicting host-virus associations; host / reservoir / vector identification; predicting zoonotic potential; predicting viral sharing; analyzing viral host range and host viral richness. Plus, some odd ones out (e.g., viral transmissibility)
With machine learning and network science, we can start to recover the source code of the global virome. We wrote an instruction manual - out now in @NatureMicrobiol - and along the way, we've tried to solve what it would mean to really "predict and prevent the next pandemic" 🧵
Practically every big question about viral ecology, evolution, and emergence - from "why do bats have so many deadly viruses?" to "can we spot a pandemic flu before the first human case?" - is a variation on a fundamental scientific challenge: predicting the host-virus network.
Over three years of research, we've compiled these kinds of studies into a unified framework, allowing us to put our finger on a new convergence science - "the science of the host-virus network" - that uses computational inference to understand viral biology across scales.
We're learning today that alphacoronavirus 1, previously not known to be zoonotic, jumped from dogs to humans > a year ago. Key lessons for where COVID-19 has been pointing us the wrong direction 🧵
1⃣ Singular focus on wildlife trade / wildlife farming as a human-animal interface is a mistake, given other natural pathways of emergence like, here, pet dogs or cats (probably).
(This doesn't mean we have to start getting rid of pets / livestock to prevent pandemics! The point of building strong healthcare systems - including One Health monitoring systems that include vets - is to stay safe by catching these kinds of events early and often.)
Something missing in a lot of viral ecology / "stop pandemics at the source" work right now:
If you're not including flu in your schema for pandemic risk, you're not actually talking about pandemics. You're talking about general disease emergence, not pandemic preparedness.
Before COVID-19, the answer to the question "what's the next pandemic most likely to be" was influenza. After COVID-19? It's actually still influenza believe it or not
So much of how we respond when a new virus emerges in a new pathway is to try to hyperfocus on sealing that entryway. But it's a bit like only locking the specific window a burglar came into your house through, and not checking the front door.