The Atlas
Nine plates. Eight pathogens the collection tracks by name, and one it cannot draw because it has not been found yet. Each plate carries a specimen sheet compiled from the published record, setting out what the thing is, the shape it takes, how much of it there is to sequence, the routes it travels and what it does to the people who catch it.
The sheets are the point. A pathogen is easy to describe in the abstract and hard to hold in the head, and most of what separates one outbreak response from another is on those lines: whether a reservoir is known, whether anything is licensed against it, how long the incubation runs, how many die. The reading sits folded beneath each sheet. Open one for why the pathogen earns a plate at all.
Every specimen is engraved from the same sampler that draws the frontispiece: the hatching follows the contour at the edges and settles to the pathogen's own angle within, so no two are struck alike. They are stylised for the plate rather than drawn to scale. The recorded dimensions carry the scale.
Ebola virus
Bundibugyo (BDBV)
- Family
- Filoviridae
- Genus
- Orthoebolavirus
- Species
- Orthoebolavirus bundibugyoense
- Form
- Filamentous, U- and 6-shaped
- Size
- 80 nm × up to 14,000 nm
- Genome
- ≈19 kb, single-stranded negative-sense RNA
- Reservoir
- Unidentified; bats suspected
- Transmission
- Direct contact with bodily fluids
- Incubation
- 2–21 days, typically 4–10
- Fatality
- 25–51% across recorded outbreaks
- Containment
- BSL-4
- Described
- 2007, Bundibugyo District, Uganda
- Held
- 2,432 articles
- Record
- The outbreak, day by day
- Reference
- Wikipedia
The plate that drove everything else. This project began as an archive of a single named outbreak, and the Bundibugyo epidemic remains by a wide margin the best-covered event in the collection.
Why it is here
There is no single Ebola virus. Six species sit in the genus Orthoebolavirus, and a species is declared only when a virus differs from its nearest relative by 30% or more of its genome. Bundibugyo shares 58 to 61% of its nucleotides with the others: close enough to produce an indistinguishable disease, far enough that little built against Zaire ebolavirus can be assumed to work against it. It was identified in western Uganda in 2007 and took four months to name. Before 2026 it had surfaced only twice, killing 37 of 149 cases in that first Ugandan outbreak and 29 of 57 in the DRC five years later. No reservoir has ever been found for it.
That taxonomic distance is the whole problem. Every licensed Ebola countermeasure, the Ervebo vaccine and the Inmazeb and Ebanga antibody treatments alike, is approved against Zaire ebolavirus alone, so the 2026 response opened with a stocked warehouse holding nothing cleared for this species; trials of remdesivir, MBP-134 and Ervebo itself began only in July. The epidemic had been running in Mongbwalu since roughly February before a laboratory flagged a filovirus that was not Zaire on 14 May, and the WHO declared a public health emergency of international concern two days after that. By 7 August it stood at 4,315 cases and 1,962 deaths, a fatality rate of 45.4%, the second largest Ebola outbreak on record and the fastest growing ever measured.
Screened on
- ebola
- ebolavirus
- EBOV
- SUDV
- BDBV
- Bundibugyo
- Ervebo
- ring vaccination
Influenza A
H5N1 (HPAI)
- Family
- Orthomyxoviridae
- Genus
- Alphainfluenzavirus
- Species
- Influenza A virus
- Form
- Spherical to pleomorphic
- Size
- 80–120 nm diameter
- Genome
- ≈13.5 kb across eight segments, negative-sense RNA
- Reservoir
- Wild waterfowl
- Transmission
- Contact with infected birds; mammalian spillover under watch
- Fatality
- ≈48% of confirmed human cases since 2003
- Described
- 1959 in birds, Scotland; 1997 in humans, Hong Kong
- Held
- 224 articles
- Reference
- Wikipedia
The second-largest holding, and the one most watched for a change in character rather than a change in volume.
Why it is here
Influenza A carries its genome in eight separate segments, and that is the whole reason it sits on this list. When two strains meet in the same host, segments swap wholesale, so an avian virus can acquire in a single step the traits it needs to move between people. There are 18 known haemagglutinin types and 11 neuraminidase types to draw from, and wild waterfowl carry most of them without falling ill. H5N1 was first found in Scottish poultry in 1959 and reached humans in Hong Kong in 1997, infecting 18 people and killing six. Between 2003 and December 2025 the WHO recorded 993 confirmed human cases and 477 deaths, a fatality rate near 48%.
What is being watched now is not the human count but the host range. The clade 2.3.4.4b lineage that emerged after 2020 has driven a global panzootic in birds and crossed repeatedly into mammals, reaching dairy cattle in March 2024 and at least 43 mammal species by June 2026. Every mammalian host is another chance at the reassortment the segmented genome makes cheap. Human vaccines exist, and the neuraminidase inhibitors oseltamivir and zanamivir still work, which is why a rise in poultry outbreaks and a rise in mammalian detections are read here as two different events.
Screened on
- H5N1
- H5N2
- H5N8
- H5N6
- H7N9
- HPAI
- highly pathogenic avian influenza
- poultry cull
Monkeypox virus
Clade I / Clade Ib
- Family
- Poxviridae
- Genus
- Orthopoxvirus
- Species
- Orthopoxvirus monkeypox
- Form
- Brick-shaped
- Size
- 220 × 300 nm
- Genome
- 190–200 kbp, double-stranded DNA
- Reservoir
- Rodents (suspected)
- Transmission
- Close contact; respiratory droplets
- Incubation
- 3–17 days
- Fatality
- 0.16% clade II; ≈5% clade Ib
- Described
- 1958, Copenhagen; first human case 1970, DRC
- Held
- 50 articles
- Reference
- Wikipedia
The odd one out on morphology: poxviruses are large enough to sit at the edge of light microscopy, and brick-shaped rather than spherical or filamentous.
Why it is here
Mpox is an orthopoxvirus, a close relative of smallpox, and it carries a double-stranded DNA genome of 190 to 200 kilobase pairs coding for some 191 proteins. That is an order of magnitude more genetic material than any RNA virus on this atlas, and it mutates far more slowly for it. The virus was found in 1958 in a Copenhagen laboratory colony of monkeys, which gave it a name that has never fitted, and the first human case was recorded in the DRC in 1970. Rodents rather than primates are the suspected reservoir.
The clade structure is what keeps it a live surveillance question rather than a settled one. Clade IIb drove the global outbreak of 2022, 93,327 cases at a fatality rate of 0.16%, and that emergency was declared over in May 2023. Clade Ib, spreading in the DRC and its neighbours since 2023, kills at roughly 5%, thirty times the rate, and prompted a second public health emergency of international concern on 14 August 2024. The countermeasures are inherited from smallpox: the MVA-BN vaccine works, ACAM2000 was approved for mpox in August 2024, and the antiviral tecovirimat failed to shorten lesion duration in clade I trials despite being approved for severe disease.
Screened on
- mpox
- monkeypox
- MPXV
- clade Ib
- clade I mpox
- orthopoxvirus outbreak
Lassa virus
Lassa, CCHF
- Family
- Arenaviridae
- Genus
- Mammarenavirus
- Species
- Mammarenavirus lassaense
- Form
- Pleomorphic, granular
- Size
- 60–300 nm diameter
- Genome
- Bisegmented ambisense single-stranded RNA
- Reservoir
- Multimammate rat
- Transmission
- Rodent excreta; nosocomial spread
- Incubation
- 7–21 days
- Fatality
- ≈1% overall; 15–20% once hospitalised
- Containment
- BSL-4
- Described
- 1969, Lassa, Borno State, Nigeria
- Held
- 34 articles
- Reference
- Wikipedia
This plate covers the viral haemorrhagic fevers that are not filoviruses: Lassa principally, with Crimean-Congo and Rift Valley fever also tracked. The arenaviruses take their name from the granular appearance of host ribosomes taken up during budding, arena, sand.
Why it is here
Lassa is the quiet one. It infects 300,000 to 500,000 people a year across West Africa and kills around 5,000, which makes it more common than every filovirus outbreak in history combined, and it draws a fraction of the attention. Most infections are mild or silent, so the overall fatality rate sits near 1%, but among those sick enough to reach a hospital it is 15 to 20%, and above 80% in the third trimester of pregnancy. A quarter of survivors are left with hearing loss. It was identified in 1969 at Lassa, in Borno State, Nigeria, after a missionary nurse fell ill. There is still no licensed vaccine, and the evidence for ribavirin, the standard treatment, is weak enough that it may worsen some cases.
Putting several agents on one plate follows the response system rather than the taxonomy. Lassa is an arenavirus carried by a rat, Crimean-Congo is a tick-borne nairovirus, Rift Valley fever is mosquito-borne, and genetically they share almost nothing. What they share is a clinical presentation, a containment protocol, and the same institutions issuing the same kind of notification, which is closer to what the collection actually screens on. The grouping helps while the record is thin; it will obscure things if any one of them becomes a story of its own.
Screened on
- Lassa
- CCHF
- Crimean-Congo hemorrhagic fever
- Rift Valley fever
- viral hemorrhagic fever
Marburg virus
MARV
- Family
- Filoviridae
- Genus
- Marburgvirus
- Species
- Marburg marburgvirus
- Form
- Filamentous, U- and 6-shaped
- Size
- 80 nm × 795–828 nm typical
- Genome
- ≈19 kb, single-stranded negative-sense RNA
- Reservoir
- Egyptian fruit bat
- Transmission
- Direct contact with bodily fluids
- Fatality
- 22–90% across outbreaks
- Containment
- BSL-4
- Described
- 1967, Marburg and Frankfurt, Germany
- Held
- 26 articles
- Reference
- Wikipedia
The other filovirus. Shares a plate family with Ebola and much of its clinical picture, but has its own reservoir, its own outbreak history, and no licensed vaccine.
Why it is here
Marburg was the first filovirus anyone met. In 1967 laboratory workers in Marburg and Frankfurt, and in Belgrade, fell ill after handling tissue from grivet monkeys imported from Uganda: 31 cases, 7 deaths, and an entirely new family of viruses. It is structurally almost indistinguishable from Ebola, the same 80 nm filament, the same 19 kb genome, the same seven genes, and it produces much the same disease. The difference that matters is that Marburg’s reservoir is known. The Egyptian fruit bat carries it, which means the spillover geography can be anticipated in a way that Ebola’s still cannot.
It is tracked separately because it behaves differently as an event. Marburg recurs in small, sharp outbreaks rather than large ones, and the fatality rate swings on how quickly they are found: 90% in Angola across 2004 and 2005, which killed 329 of 374, against 44% in Equatorial Guinea in 2023, 67% in Tanzania the same year, and 22% in Rwanda in 2024. Nothing is licensed against it. Early rehydration and supportive care are the whole of the response, which is why detection speed shows up so plainly in those numbers.
Screened on
- Marburg
- MARV
- MVD
Nipah virus
NiV
- Family
- Paramyxoviridae
- Genus
- Henipavirus
- Species
- Henipavirus nipahense
- Form
- Pleomorphic
- Size
- 40–600 nm
- Genome
- ≈18 kb, non-segmented negative-sense RNA
- Reservoir
- Fruit bats
- Transmission
- Contaminated date palm sap; close contact
- Incubation
- 4–14 days, up to 45
- Fatality
- 40–75%
- Containment
- BSL-4
- Described
- 1998, Sungai Nipah, Malaysia
- Held
- 11 articles
- Reference
- Wikipedia
A small holding for a pathogen that sits high on every priority list: high case fatality, no licensed vaccine, and a transmission route that puts it in contact with dense populations.
Why it is here
Nipah emerged in 1998 among pig farmers in peninsular Malaysia and took its name from the village of Sungai Nipah. That first outbreak ran to 265 cases and 108 deaths. Pteropus fruit bats are the reservoir, and the virus reaches people by several routes at once: through livestock, from person to person, and through raw date palm sap collected overnight in pots that bats drink from. The fatality rate runs between 40 and 75%. Incubation is usually 4 to 14 days but has been recorded at 45. No vaccine is licensed, though ChAdOx1 NipahB entered Phase II trials in December 2025 and the monoclonal antibody m102.4 is in Phase 1.
The outbreaks are small and they keep happening. Bangladesh sees them most winters, tied to the sap harvest; Kerala has had repeated introductions, including deaths in 2021 and 2024, and confirmed cases appeared in West Bengal in January 2026. Eleven articles is the thinnest holding on the atlas, and it measures the size of the outbreaks rather than their importance. A pathogen that kills more than half the people it infects and has nothing licensed against it is not a minor entry. It is one that generates little press until the year it does.
Screened on
- Nipah
- NiV
- henipavirus
- Nipah encephalitis
Prion
PrP^Sc
- Family
- Not applicable
- Form
- Amyloid fibril
- Size
- ~10 nm fibril
- Genome
- None; no nucleic acid of any kind
- Reservoir
- Not applicable
- Transmission
- Ingestion; iatrogenic; sporadic; inherited
- Incubation
- 5 to 20 years or more
- Fatality
- Invariably fatal
- Described
- 1982, named by Stanley Prusiner
- Held
- 2 articles
- Reference
- Wikipedia
Not a virus. Not an organism at all, but a misfolded protein that propagates its own conformation, with no genome to sequence and no reservoir in the ordinary sense. It is on the plate list because it is infectious, and because the category “pathogen” has to be able to hold things that do not behave like the others.
Why it is here
A prion is the protein PrP folded wrongly. The infectious form, PrPSc, carries far more beta sheet than the healthy cellular form, and it converts normal copies to its own shape on contact, stacking them into amyloid fibrils until the cell dies. There is no nucleic acid anywhere in it. Stanley Prusiner named the thing in 1982 and took the Nobel Prize in 1997 for showing that a protein alone could transmit disease. In people the result is Creutzfeldt-Jakob disease in its sporadic, familial, iatrogenic and variant forms, along with kuru, Gerstmann-Straussler-Scheinker syndrome and fatal familial insomnia. In animals it is BSE, scrapie, and chronic wasting disease in deer and elk.
Prion disease defeats nearly every instrument built for the rest of this atlas. There is no genome, so genomic surveillance has nothing to read. Incubation runs 5 to 20 years or more, which separates an exposure from its outbreak by decades. The agent resists proteases, heat, ionising radiation and formaldehyde, and ordinary sterilisation does not touch it; decontamination needs sodium hydroxide or hypochlorite. Every case is fatal and none is treatable. Two articles is the smallest holding here, and it stands as a deliberate test of whether the collection notices things it is not looking hard for.
Screened on
- prion
- CJD
- vCJD
- BSE
- scrapie
- Creutzfeldt-Jakob
- chronic wasting disease
Plague
Yersinia pestis
- Family
- Yersiniaceae
- Genus
- Yersinia
- Species
- Yersinia pestis
- Form
- Short rod, bipolar staining
- Size
- 0.5–0.8 µm × 1–3 µm
- Genome
- ≈4.6 Mb chromosome plus three plasmids
- Reservoir
- Wild rodents, carried between them by fleas
- Transmission
- Flea bite (bubonic); respiratory droplets, person to person (pneumonic)
- Incubation
- 2–8 days bubonic; 1–3 days pneumonic
- Fatality
- 30–60% bubonic untreated; pneumonic near 100% untreated
- Containment
- BSL-3
- Described
- 1894, Hong Kong
- Held
- 602 articles
- Reference
- Wikipedia
The only bacterium in the atlas, and the oldest name on it. A focus of the collection since October 2026, on the same footing as Ebola.
Why it is here
Plague never went away. Yersinia pestis lives in wild rodent colonies across Central Asia, Africa and the Americas and reaches people a few hundred times a year, mostly in Madagascar and the Democratic Republic of the Congo. The common form is bubonic, carried by flea bite and treatable with ordinary antibiotics if caught early. The form that matters is pneumonic: an infection of the lungs that spreads from person to person through the air, kills within days, and is close to uniformly fatal without treatment. It is what turned the 1910 Manchurian epidemic into some 60,000 deaths, and what carried Madagascar’s 2017 outbreak past 2,400 cases, most of them pneumonic.
On the night of 1 October 2026 a 28-year-old researcher at the Irkutsk Anti-Plague Research Institute of Siberia and the Far East died in hospital in Shelekhov, three days after being admitted. The head of neighbouring Buryatia referred to her as the woman who had died of plague, then edited his post to say she “may have”. Rospotrebnadzor, Russia’s health watchdog, called it pneumonia of unknown aetiology and told the WHO that no plague case had been recorded in Irkutsk. The Shelekhov hospital was quarantined, around 200 people were placed under observation, and reports of the death began disappearing from Russian news sites. By 6 October the United States was pressing Moscow for information and reports of a second death were circulating unconfirmed.
The institute holds live plague strains as part of the network Russia maintains to watch its own natural foci, among them the ones along the Mongolian border from Altai to Transbaikalia. Whether she died of plague, and if so how she was exposed, can so far be answered only from inside Russia, and the public account has already changed once. The collection keeps every version of it, the official ones included, as it is published and before it is revised.
Screened on
- plague
- Yersinia pestis
- pneumonic plague
- anti-plague
- Irkutsk
- чума
Pathogen X
There is no specimen for this plate, so the engraving never settles. The contour does not close, the hatching will not lie down, and the second impression stays out of register. It is a drawing of something the engraver has not seen.
A serious international epidemic caused by a pathogen currently unknown to cause human disease.
Why it is here
That is WHO’s own designation. Disease X was added to the R&D Blueprint priority list in February 2018, alongside Ebola, Zika, SARS, Lassa fever, Marburg and Nipah, on the reasoning that preparedness has to be built for a class of threat rather than for a named member of it. It is not a metaphor but a real placeholder on a real list, put there because the pathogen that matters most is reliably the one without a name yet. Most observers consider COVID-19 to have been the first of them. “In a nutshell, Covid-19 is Disease X,” as Peter Daszak put it in February 2020.
That is also the problem this collection was rebuilt around. Every recent pandemic and near miss was, in its earliest and most valuable reporting window, an unnamed cluster: COVID was viral pneumonia of unknown cause, MERS was a novel coronavirus, H5N1 in mammals began as unexplained die-offs. Naming happens after the reporting most worth keeping. An archive that screens on disease names is therefore structurally guaranteed to miss the lead-up signal it exists to capture, which is why the pipeline now carries a syndromic catch-all and lets authoritative sources through ungated.