Nipah Virus at the Neuroimmune Interface

Nipah virus is usually framed as a high-fatality zoonotic threat. But its neurological heterogeneity raises a deeper question: how do viral pathogenesis, host immunity and neuroinflammation interact to determine who develops severe encephalitis — and who recovers?

Nipah virus occupies an unusual position among emerging infectious diseases. It is a zoonotic paramyxovirus maintained naturally in fruit bats, capable of transmission through intermediate animal hosts, contaminated food and direct human contact, and able to cause severe respiratory and neurological disease. Its estimated case-fatality rate varies substantially between outbreaks but is commonly reported between 40% and 75%, and there is currently no approved human vaccine or specific licensed antiviral treatment.¹ Those characteristics alone justify its place on the World Health Organization’s Research and Development Blueprint list of epidemic threats requiring urgent research.²

Yet Nipah presents another scientific problem that deserves particular attention: why does infection produce such profoundly different neurological trajectories between individuals? Some infections remain relatively limited, while others progress rapidly to encephalitis, seizures, impaired consciousness and critical illness. Among survivors, neurological recovery may be incomplete, and relapsing or late-onset encephalitis has been documented months or even years after apparent recovery.^3,4,5

Understanding that heterogeneity requires more than virology alone. It requires investigation of the interaction between viral tropism, vascular injury, innate antiviral immunity, inflammatory responses, host susceptibility and the central nervous system — which makes it a neuroimmunological problem as well as an infectious-disease one. And because Nipah outbreaks are comparatively uncommon, geographically dispersed and composed of small patient cohorts, it is also a particularly challenging health-data problem.

A zoonotic virus with a neurological phenotype

Nipah virus was identified following an outbreak of encephalitis among pig farmers in Malaysia in 1998–1999, an event that established an important early One Health pathway: Pteropus fruit bats were the natural reservoir, pigs acted as amplifying hosts, and occupational contact created an effective route into humans.^6,7

The epidemiology observed subsequently in Bangladesh and India has been different. Direct or indirect bat-to-human transmission has played a greater role, particularly through consumption of raw date-palm sap contaminated by bat saliva or urine, and human-to-human transmission has also occurred within households and healthcare settings.^7,8,9

That variation matters scientifically, because “Nipah infection” is not necessarily a homogeneous exposure. Different outbreaks may involve different routes of transmission, viral strains, inoculum, healthcare environments and host populations, so treating cases collected across outbreaks as though they were interchangeable can obscure important biological and epidemiological variation. This becomes particularly consequential when attempting to identify predictors of neurological severity or outcome, since an apparent predictor may simply be a marker of which outbreak a patient belonged to.

Why does Nipah affect the brain?

The neuropathology of Nipah is unusually instructive, because human pathological studies have demonstrated systemic vasculitis, endothelial infection, thrombosis, microinfarction and direct neuronal infection, which means cerebral pathology appears to result from more than one mechanism: vascular injury can compromise cerebral perfusion and produce focal tissue damage, while the virus can also infect neurons directly.¹⁰ That dual mechanism distinguishes Nipah encephalitis from a simplistic model in which a neurotropic virus merely enters neurons and destroys them.

There is a molecular basis for both halves of that picture. Nipah uses ephrin-B2 and ephrin-B3 as cellular receptors, and these proteins are expressed both in vascular endothelial cells and in relevant neuronal populations. Contemporary reviews therefore describe Nipah pathogenesis as involving endothelial infection, systemic vasculitis, disruption of vascular and blood–brain-barrier integrity, and direct infection of the central nervous system.^11,12

The clinical consequence can be extensive: in the original Malaysian series, reduced consciousness was common among patients with Nipah encephalitis, and neurological manifestations included brainstem dysfunction, segmental myoclonus, seizures and rapidly progressive encephalopathy.³ Even so, viral neuroinvasion alone is unlikely to explain the full spectrum of disease, which is where host response enters the argument.

Nipah as a neuroimmunological disease

The immune response to a neurotropic virus must solve two competing problems simultaneously. It has to control the pathogen rapidly enough to prevent uncontrolled viral replication and dissemination, while keeping inflammatory responses within the central nervous system sufficiently regulated to avoid causing additional tissue injury. Nipah complicates this further, because henipaviruses possess mechanisms capable of interfering with host antiviral immunity: viral proteins can inhibit type-I interferon signalling and other components of innate antiviral defence.¹³

That creates an important conceptual distinction, because the immunological question is not whether a stronger immune response is good or bad, but which immune responses provide effective viral control, at what stage of infection, and when immune activation begins to contribute to pathology rather than protection. The answer is unlikely to be captured by any single inflammatory biomarker, because innate antiviral signalling, endothelial activation, immune-cell recruitment, cytokine production, adaptive T-cell responses and antibody responses all operate across different timescales. The biological significance of an inflammatory signal may also depend on whether it occurs before invasion of the central nervous system, during active viral replication, or during later tissue recovery.

Recent evidence adds another dimension to that picture, in that Nipah-specific T-cell memory has been demonstrated in Malaysian survivors more than two decades after natural infection, showing that the virus can generate remarkably durable adaptive immunity in those who recover.¹⁴ The challenge, then, is to move from describing “inflammation” as a single phenomenon towards understanding immune trajectories.

From a biomarker to a trajectory

Clinical studies often measure biological variables at a single time point, and for a disease like Nipah that approach may be inadequate. Consider two hypothetical patients who enter hospital with similar fever, viral burden and C-reactive protein concentrations. Over the following days their biology diverges: in one, viral replication comes under control, inflammatory activity resolves and neurological function remains preserved, while in the other, antiviral control fails or inflammatory activity becomes dysregulated, vascular and central-nervous-system involvement progresses, and encephalitis develops. A single admission measurement cannot describe that process, whereas a longitudinal dataset potentially can.

This is why an important direction for Nipah research is not simply the identification of biomarkers associated with mortality, but the characterisation of biological trajectories associated with neurological progression and recovery. Such trajectories might eventually incorporate:

• viral load and viral kinetics;
• innate antiviral signalling;
• inflammatory mediators;
• routine haematological measures;
• markers of endothelial injury;
• cerebrospinal-fluid findings;
• neurological examination and level of consciousness;
• neuroimaging;
• respiratory involvement;
• treatment and supportive-care variables;
• and longitudinal functional outcomes.

The critical question is whether combinations and trajectories of these observations explain more than any one measurement alone. That is an inherently multivariate and longitudinal problem, and it is precisely where health data science can contribute.

The problem of small data

Emerging infectious disease research frequently encounters a paradox: the diseases for which better predictive models would be most valuable are often the diseases for which the least data exist, and Nipah illustrates that clearly. Large machine-learning models are attractive when tens or hundreds of thousands of well-characterised observations are available, and Nipah produces a fundamentally different statistical environment — relatively small outbreaks, missing observations, changing clinical protocols, heterogeneous populations and few severe neurological outcomes from which to learn.

The appropriate response is not to abandon quantitative modelling but to use methods suited to the available evidence. For Nipah, that points towards Bayesian analysis, hierarchical or multilevel modelling, longitudinal mixed models, survival analysis, causal inference and careful external validation between outbreaks. Hierarchical approaches are especially attractive conceptually, because they allow observations from individual outbreaks to inform a broader model while preserving the differences between geographic locations, transmission settings and patient populations, rather than collapsing every patient into one undifferentiated dataset. Bayesian methods, similarly, can incorporate established biological knowledge while representing uncertainty explicitly, which matters a great deal when sample sizes are small. The methodological objective should therefore not be maximum model complexity, but maximum inference from limited data without overstating certainty.

Data quality comes before artificial intelligence

There is a temptation, whenever emerging infections are discussed, to move immediately towards artificial intelligence. For Nipah that risks beginning at the wrong end of the problem, because a sophisticated algorithm cannot compensate for poorly defined outcomes, incompatible variables, uncertain provenance or inconsistent sampling schedules.

Before useful predictive modelling can occur, researchers need to know what individual data points actually represent. Was an inflammatory marker taken on the first, third or tenth day of illness? Was encephalitis diagnosed clinically, radiologically or using a standardised research definition? Was neurological outcome measured at discharge, at thirty days, at six months or several years later? Did different centres use identical laboratory assays? Was the patient transferred between hospitals, and was any observation corrected subsequently? How was infection confirmed?

These are not administrative details; they determine whether datasets can legitimately be compared at all. WHO’s Nipah R&D roadmap has long identified the need for coordinated research and accelerated development of diagnostics, therapeutics and vaccines, and current WHO readiness guidance likewise emphasises surveillance, laboratory capacity, clinical management, coordination and multisectoral assessment.^15,16 For analytical research, those priorities imply a further requirement: data harmonisation should begin before the next outbreak rather than after it.

Provenance is a scientific requirement

Reliable emerging-disease analysis requires not simply data integration but traceable integration. If a clinical observation, sequencing result or immunological measurement contributes to a model, investigators need to understand its origin — who generated it, where it originated, which method produced it, when it was collected, whether it was transformed, whether it was subsequently amended, and which analytical version used it.

This matters most precisely where it is hardest, which is when datasets cross institutional or disciplinary boundaries. A Nipah research dataset might eventually incorporate clinical records, diagnostic laboratory results, sequencing, wildlife surveillance, veterinary information and environmental observations, each with different methods, uncertainties and governance requirements. Interoperability without provenance can create an illusion of integration while concealing incompatibility. The objective should therefore be FAIR and reproducible data that remain interpretable in their original biological and clinical context, rather than merely larger repositories — a problem we have written about more generally in Biological Representation Is Not a Data-Formatting Problem.

Connecting the patient to the ecosystem

The patient with Nipah encephalitis represents only one point in a longer biological system. The complete chain runs from bat ecology through viral shedding, environmental or food contamination, livestock or direct human exposure, infection, human transmission, within-host viral replication, host immune response and neuroinvasion to encephalitis, and finally to recovery, disability or death.

Different research disciplines traditionally observe different portions of that chain: ecologists study bat populations, veterinary researchers investigate animal infection, epidemiologists reconstruct transmission, virologists study viral variation, immunologists examine host defence, neurologists characterise disease of the central nervous system, and data scientists analyse patterns across observations. The scientific opportunity lies in joining the appropriate layers without pretending they are equivalent. The One Health approach advocated by WHO, FAO and WOAH reflects exactly this need to consider human, animal and environmental evidence together when addressing zoonotic threats.^17,18 For a pathogen such as Nipah, integration is not simply desirable organisational behaviour; it reflects the biology of the disease. Our sister publication One Health Security examines that systems argument in detail in Nipah Virus: When a Bat Virus Reaches the Brain.

Neurological recovery deserves greater attention

The acute mortality associated with Nipah understandably dominates discussion of the disease, but survival does not necessarily represent the endpoint. WHO notes neurological sequelae in approximately one in five survivors, and historical cohorts have documented relapsing and late-onset encephalitis months or even years after apparent recovery.^1,4 More recent systematic analysis of survivors continues to identify long-term neurological and functional consequences, although the available evidence remains limited by small cohorts and uneven follow-up.⁵

That makes longitudinal follow-up scientifically important, and it shifts the question beyond whether the patient survived. How completely did neurological function recover? Which cognitive, motor, behavioural or fatigue-related symptoms persisted? Were neurological outcomes associated with characteristics of the acute immune response? And does residual or recurrent neurological disease reflect persistent viral biology, immune-mediated injury, structural damage, or some combination of the three? These are difficult questions, and they are also precisely the questions that become answerable only when acute clinical data are linked reliably to long-term outcomes.

Current events underline the relevance

Nipah remains a contemporary health threat rather than a historical one. In January 2026, India notified WHO of two laboratory-confirmed infections in healthcare workers at a private hospital in Barasat, West Bengal; both experienced severe disease including neurological illness, and extensive contact investigation identified more than 190 contacts, all of whom tested negative. That event was the third Nipah outbreak reported in West Bengal, following Siliguri in 2001 and Nadia in 2007.¹⁹ In June 2026, Kerala reported a further laboratory-confirmed infection in Kozhikode district; the patient presented predominantly with neurological manifestations and required intensive-care ventilatory support, and 104 contacts were identified for active monitoring with no secondary cases reported at the time of WHO’s update.²⁰

These events illustrate an important feature of Nipah preparedness. Modern outbreak response can contain transmission remarkably effectively through rapid diagnosis, isolation, contact tracing, laboratory support and surveillance. Yet every sporadic spillover also represents an opportunity to learn, and when cases are this rare, losing well-characterised clinical, virological, immunological or longitudinal data from even a small number of patients is a material scientific cost.

A research agenda at the intersection of infection, immunity and data

Nipah should therefore be considered not only an emerging zoonotic threat but an important model system for studying the relationship between infection and neurological injury, and several research questions stand out from that framing.

What determines neuroinvasion? Understanding why certain patients develop disease of the central nervous system while others do not requires simultaneous consideration of viral characteristics and host susceptibility, rather than either in isolation.

Which immune responses are protective? Effective innate and adaptive responses are necessary for viral control, but the timing and magnitude of those responses may matter as much as their presence.

Which responses contribute to neurological injury? Vascular inflammation, endothelial dysfunction, immune activation within the central nervous system and direct neuronal infection may interact rather than operate independently.

Can early biological trajectories predict neurological deterioration? Repeated measurements may reveal clinically useful patterns that are simply invisible in cross-sectional data.

Why do neurological outcomes differ among survivors? Long-term cohort data are needed to connect acute infection and immune response to subsequent cognitive and neurological function.

Can evidence be generalised across outbreaks? Models should be tested between independent populations rather than judged solely by how accurately they reproduce the dataset from which they were developed.

These questions require collaboration between virology, immunology, neurology, veterinary science, epidemiology and quantitative health research, and none of those disciplines is sufficient in isolation.

Preparing the data before the emergency

Nipah’s rarity makes preparedness unusually important, because when an outbreak begins there is little time to negotiate common outcome definitions, design data structures, create analytical protocols or determine how specimens and clinical observations should be linked — yet all of those decisions can be made beforehand. A mature research infrastructure for emerging neuroinfectious disease could define minimum clinical datasets, common neurological outcomes, standard sampling time points, core immunological and virological variables, long-term follow-up measures, data provenance requirements and pre-specified analytical frameworks.

This would not dictate what individual researchers study. It would make findings generated by different centres substantially easier to compare and combine. The principle also extends well beyond Nipah, because Japanese encephalitis, West Nile neuroinvasive disease, tick-borne encephalitis, rabies and future zoonotic neurological pathogens present related problems: heterogeneous exposure, relatively uncommon severe neurological outcomes, immune-mediated pathology and fragmented datasets. Nipah can therefore serve as a model for a wider methodological challenge, and it is the same design problem that motivates our work on Sybil.

From fragmented observations to evidence

The defining difficulty of emerging infections is often not a complete absence of information but its fragmentation. One laboratory holds genomic data and another holds immunological measurements; a hospital holds the longitudinal clinical record; veterinary services hold animal-health observations; wildlife researchers understand reservoir ecology; and public-health teams hold contact and exposure information. Each dataset answers only part of the question.

The future of emerging-disease research will therefore depend not only on generating more information but on making evidence interoperable, traceable, analytically rigorous and biologically interpretable, and Nipah virus demonstrates why that matters: it begins as an ecological and zoonotic event, becomes an infectious-disease problem, and in some patients becomes a neurovascular and neurological catastrophe. Behind those transitions sits the host immune response: essential for viral control, potentially implicated in pathology, and still incompletely understood.

Understanding that journey will require us to connect biological mechanisms with longitudinal clinical observations, and to extract reliable inference from datasets that will, by their nature, remain comparatively small. That is not a reason to expect less from data science. It is a reason to practise it more carefully.