Mapping the global diversity and surveillance landscape of medically important bat-borne viruses: a scoping review

¹Department of Biochemistry, Jomo Kenyatta University of Agriculture and Technology, Nairobi City, Kenya, ²Centre for Virus Research, Kenya Medical Research Institute, Nairobi City, Kenya, ³Department of Entomology and Vector Borne Infections, Walter Reed Army Institute of Research- Kisumu City, Kenya, ⁴Institute for Biotechnology Research, Jomo Kenyatta University of Agriculture and Technology, Nairobi City, Kenya, ⁵Research Department, Zihi Institute, Nairobi City, Kenya, ⁶Pwani University Bioscience Research Centre (PUBReC), Pwani University, Kilifi, Kenya

^&Corresponding author
George Kiprono Kisoi, Department of Biochemistry, Jomo Kenyatta University of Agriculture and Technology, Nairobi City, Kenya

Introduction    

Bats are one of the most ecologically diverse and widely distributed mammalian groups with over 1,500 species [1,2]. With the exception of Antarctica, they live on almost every continent and occupy a variety of ecological niches. This variation in species, diet, and roosting habits shapes opportunities for viral maintenance and transmission by enabling wide-ranging contact networks both within and between bat populations and other animals [3,4]. Viruses of public health significance, such as coronaviruses (Coronaviridae), filoviruses (Filoviridae), henipaviruses (Paramyxoviridae), and lyssaviruses (Rhabdoviridae), are found in several bat families, including Pteropodidae, Rhinolophidae, Vespertilionidae, and Phyllostomidae [5-9].

Bats' extended lifespan, colonial behaviors, and powered flight are factors that contribute to special viral relationships and the capacity to transmit viruses across species [10-16]. They are able to host numerous viruses without showing any clinical disease, an aspect attributed to flight-adaptive immune systems [17,18]. This permits viral persistence and shedding, which can cause spillover to other animals and humans [19], although the effect of these viruses on bat's health is not well understood [20-23]. This is demonstrated by a well-documented finding that bats are natural reservoirs of Severe Acute Respiratory Syndrome (SARS)-like coronaviruses [24].

By mapping the viral diversity of bat families and regions, one can understand the ecological and evolutionary processes that cause spillover risk [25]. However, global knowledge of bat viral diversity is lacking, with sampling concentrated in certain regions. While Asia and Africa have been relatively well sampled, Oceania and parts of South America have received less research attention [25-27]. We note that 'well sampled' and 'less sampled' reflect relative research effort, not an absolute standard of adequacy. This pattern is based on counts of sequence deposition (section 3.2) rather than a claim of research attention because dense sequence counts can still originate from a small number of sampling sites within a given region, and sampling that did not yield positive detections is rarely published or deposited. Therefore, true surveillance effort, and its geographic unevenness, likely underestimated by these figures. Also, this should not be read as a blanket statement regarding bat virus research attention in every continent or region or even a virus group. For instance, Oceania has a history of bat lyssavirus and henipavirus surveillance in Australia, and Africa remains understudied for bat viruses in general considering beyond filovirus surveillance [28]. Also, the concept of the zoonotic potential is inconsistently used in literature and is often based on phylogenetic relatedness, which can overestimate the actual spillover risk [29-31].

Databases such as Database of Bat-associated Viruses (DBatVir) (now migrated to ZOVER) [32] catalog bat-virus associations but primarily provide raw sequence data without standardized risk classifications [33]). The most recent iteration, ZOVER 2.0, has expanded to include virome level data and enhanced search functionalities [34]. This review maps and quantifies global bat-virus surveillance from January 2010 to November 2024. We aimed to investigate taxonomic and geographic differential sampling intensities in surveillance, determine viral richness by bat host family and geographic region, assess the reliability and consistency of zoonotic risks labels and propose a common system to measure spillover potential. This review aligns with the One Health paradigm [35-37], which acknowledges that anthropogenic processes increase the frequency of interactions among bats, livestock, and humans [38-40].

Methods     

Protocol registration: the review protocol was registered on the Open Science Framework (OSF) on 16 October 2024, in accordance with PRISMA-ScR recommendations [41].

Study design

We screened peer-reviewed articles, preprints, and database entries from PubMed, Google Scholar, BioOne, NCBI GenBank, ZOVER, and DBatVir before its migration to identify published work on viral pathogens associated with bats. Two reviewers independently screened titles/abstracts and full texts and resolved disagreements via consensus (GK and JN); a third reviewer (SL) arbitrated unresolved conflicts. After screening, a total of 175 peer-reviewed research articles were identified and subsequently analyzed based on the study objectives. PRISMA guidelines were followed for the selection and eligibility identification of study materials [42].

Eligibility criteria

Inclusion and exclusion criteria

We included peer-reviewed articles, preprints, and database entries that reported novel or known virus detection in bat species. We excluded studies focusing solely on serology without molecular identification and studies not published in English. The PRISMA 2020 guideline for systematic reviews was followed (Figure 1). Preprints were included to capture the most recent surveillance data, recognizing that they have not undergone peer review; sensitivity analyses comparing preprints to peer-reviewed articles showed no systematic differences in reported virus families or geographic distributions (χ² = 2.34, df = 2, p = 0.31).

Search strategy and iterative refinement

We searched for records from January 1, 2010, to November 30, 2024. In our first search, we used specific terms like "medically important bat viruses." However, this approach missed key studies, particularly from Africa, where researchers often focus on general viral detection rather than specific important viruses. Ignoring non-English studies may also mean missing relevant information published in regional journals, including those in Chinese, Portuguese, and Spanish. We mention this limitation in the limitations section. To gather all literature on bat virus surveillance, we improved our search strategy through three cycles of testing and validation. The updated strategy used broader terms across four related areas. Complete search strings, database-specific parameters, and filters applied for each of the eight search components are summarized in Table 1 (Appendices A-H). Specifically, appendix A details the PubMed and NCBI search using MeSH terms and boolean operators. Appendix B describes the ZOVER viral database search for bat hosts and high-concern virus families. Appendix C covers BioOne searches for spillover risk and human-wildlife interfaces. Appendix D outlines Google Scholar title and text searches. Appendix E specifies National Library of Medicine MeSH terms. Appendix F describes PROQUEST full-text searches. Appendix G covers ProMED outbreak keyword searches. Appendix H presents the iteratively refined strategy including Africa-specific enhancements and validation metrics showing a 115% overall improvement in yield (447% for African studies). Filters include English language, publication years 2010-2024, and peer-reviewed status where applicable. This adjustment increased our initial yield by 115%, going from 2,147 to 4,617 total records. For studies from Africa, the increase was 447%, rising from 89 to 487 records.

Extraction and synthesis of data

Data extraction was performed independently by two authors (GK and JN) using a standardized form. Disagreements were resolved by consensus; a third author (SL) arbitrated unresolved conflicts (fewer than 5% of extractions).

Tiered zoonotic potential classification framework

We used a new evidence-based tiered system in response to the substantial and inconsistent use of the term "zoonotic" noted in the literature. This was applied to re-classify each virus record based on the most recent evidence as of November 2024. The framework categorizes viruses into four distinct tiers. Tier 4 which is confirmed zoonotic or documented human disease such as Nipah virus, and Rabies virus. Tier 3, representing high zoonotic potential, defined by experimental evidence of human cell infection (e.g., SARS-like bat coronavirus WIVI (WIV1-CoV), and swine acute diarrhea syndrome coronavirus (SADS-CoV) [43]. Tier 2 (moderate potential): phylogenetic relatives with 70-90% amino acid identity in key genomic regions (e.g., Bat Coronavirus HKU4) [44,45]. Tier 1 (low potential): <70% identity to known human pathogens or belonging to genera with no documented human infection. In essence, the framework moves from confirmed human disease (Tier 4) to experimental cell infection evidence (Tier 3), to suggestive phylogenetic relatedness (Tier 2), and finally to distant or no known relation to human pathogens (Tier 1).

Analytical approach

The curated subset (n=244) was not selected through a sample size calculation but rather, it was manually assembled across continents, virus families, and zoonotic-risk tiers, as a manageable subset for validating whether patterns in the full dataset of 17, 134 records held up under closer, record level scrutiny, including manual verification of ecological metadata that is bat diet, sampling landscape, intermediate host not extractable from full dataset. Representative of this subset relative to the full dataset was then confirmed statistically, as follows: We validated representativeness by comparing the distribution of virus families and geographic regions in the curated subset (n=244) to the full corpus or dataset (n=17,134) using chi-square goodness-of-fit tests [46]. No significant differences were found for virus family distribution (χ² = 6.72, df = 8, p = 0.57) or continental distribution (χ² = 4.18, df = 5, p = 0.52), indicating the subset was representative.

Data analysis

We used Python v3.9 and the pandas library to clean and manipulate the data, scipy.stats to perform descriptive statistics and chi-square tests, and matplotlib and seaborn to visualize the results. We defined 'viral richness' as the number of unique virus species (or strains where species-level assignment was not available) identified in each geographic region or bat host family. We used the total number of genomic sequences deposited in public databases as a proxy for surveillance effort, recognizing that sequences may not directly equate to samples or studies.

Quality assessment and uneven distribution assessment

The Database of Abstracts of Reviews of Effects (DARE) criteria assessed study synthesis adequacy, inclusion and exclusion clarity, and search comprehensiveness. Risk of publication bias was assessed visually using a funnel plot of sequence count against study sample size. Regional bias was assessed by comparing observed sequence counts per continent to expected counts based on bat species richness (Chi-square goodness-of-fit).

Results     

Study selection and characteristics of the global dataset

Our systematic search and screening process yielded a final dataset of 17,134 bat-associated virus genomic sequences, extracted from 175 included sources (Figure 1). This dataset formed the basis for all subsequent analyses. It shows the number of records identified through database searching (n = 3,598 across 12 sources). Materials removed prior to screening (n = 1,970) due to: duplication, full-text unavailability, deviation of data from the topic of study, and very old publications. Screened materials (n = 1,628) were assessed for eligibility. A total of 1,453 materials were excluded during a full text screening because they were duplicates or fell outside the scope of the study objectives. A total of n = 175 studies were ultimately included in the scoping review, contributing 17,134 viral sequences. This flow follows the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) guidelines.

Geographic uneven distributions in global surveillance efforts

Our analysis revealed an uneven distribution of surveillance effort, measured by genomic sequence deposition, across continents (Figure 2 A). Analysis of variance (ANOVA) confirmed that the number of genomic sequences deposited per continent differed significantly (F(5,170) = 15.34, p < 0.001). Asia accounted for the largest share of sequences (6,794; 39.7% of the 17,134 geographically assigned sequences), followed by Africa (5,107; 29.8%). The complete continental distribution, including percentages and total assigned sequences, is provided in Supplementary Annex 1 A. Country-level surveillance intensity is summarized in Supplementary Annex 1 B, which indicates that China, Brazil, Australia, and the United States have contributed the majority of records, while dozens of countries have only one or two detections. In contrast, North America (2,000; 11.7%), Europe (1,613; 9.4%), South America (1,371; 8.0%), and Oceania (249; 1.5%) were significantly less frequently sampled. Ninety-two sequences were excluded from geographic analysis due to missing location data. When normalized by bat species richness, the uneven distribution becomes starker (Figure 2 B), with sampling intensity indices of 0.23 for Oceania and 0.33 for South America, compared with 1.81 for Asia and 1.42 for Africa. Here, bat species richness refers to the total number of extant bat species documented as native to each continent not the number of species sampled in the included studies. The values per continent were obtained from the Mammal Diversity Database V1.12.1. Annual deposition of sequences peaked in 2023 (n = 3,500 sequences), with increases following major zoonotic outbreaks and a growing share of studies using next-generation sequencing (NGS) (Figure 3).

Taxonomic distribution of viral diversity

Nearly 79% of all virus data originates from just three bat families: Vespertilionidae (6,028 records), Rhinolophidae (3,884 records), and Pteropodidae (3,554 records) (Figure 4 A). In contrast, Emballonuridae (n = 175, 1.0%), Nycteridae (n = 51, 0.3%), and Megadermatidae (n = 16, 0.1%) were underrepresented, and 1,011 records (5.9%) came from unclassified Chiroptera. When normalized by bat species richness, research intensity is highest for Rhinolophidae (approximately 12 records per species), followed by Pteropodidae (approximately 8 records per species) and Vespertilionidae (approximately 4 records per species) (Figure 4 B). Representative zoonotic bat-borne viruses reported from Africa, Asia, Europe, South America, North America, and Oceania are listed in Supplementary Annex 1 C-G, respectively, and the number of virus records per bat family is given in Supplementary Annex 1 H. Novel virus discovery continues to expand known diversity, including recently characterized hepadnaviruses and hepeviruses in Chinese bat populations. For example, metagenomic surveys of six bat species in close contact with humans in southern China have revealed diverse viral communities including coronaviruses, astroviruses, and picornaviruses [47].

Spillover risk classification

Application of our evidence-based tiered framework to the large-scale dataset revealed that nearly three-quarters of recently discovered bat viruses possess confirmed evidence for zoonotic potential (Figure 5 A). Of the 158 bat viruses discovered between 2010 and 2024, 25% were classified as Tier 4, 32% as Tier 3, 28% as Tier 2, and 15% as Tier 1. Multi-scale validation of the tiered risk framework is shown in Supplementary Annex 2 A. The tier distribution of the curated subset (N = 244) closely matched that of the full dataset (N = 17,134), supporting its representativeness. Supplementary Annex 2 B shows the continental distribution of high-risk (Tier 4 and Tier 3) viruses from the curated dataset across six continents. Tier 4 viruses were most frequently detected in Europe (n = 14), South America (n = 11), and Asia (n = 6), and Tier 3 viruses in Asia (n = 28), Europe (n = 18), and Africa (n = 14). This distribution was validated by our curated dataset analysis (n=244 records), which showed closely aligned proportions where 74.6% of viruses with confirmed or probable zoonotic potential. A comparison between the original study labels and our evidence-based classification (Figure 5 B) revealed that original studies substantially overstated zoonotic risk, particularly for Tier 2 viruses, with a 45.6% overstatement gap (original studies labeled 73.6% of Tier 2 viruses as zoonotic, versus 28.0% in our classification). Filoviruses such as Ebola and Marburg viruses, are among the most high-concern bat-borne pathogens.

For instance, Marburg virus outbreaks have been linked to Rousettus aegyptiacus bats in sub-Saharan Africa [48]. Lyssaviruses, including rabies virus, represent significant bat-borne zoonotic pathogens and continues to evolve in bat reservoirs, with ongoing spillover risk [49]. Additional bat-borne zoonotic viruses with documented or suspected spillover potential are listed in Supplementary Annex 1 I. The influence of detection method on risk tier assignment is shown in Supplementary Annex 2 C, where PCR-based methods have unequally identified Tier-4 viruses, while NGS has uncovered more Tier 3 viruses. PCR-based methods show higher detection of Tier 4 viruses (35% vs. NGS 15%). NGS-based methods show higher detection of Tier 3 viruses (45% vs. PCR 20%) and slightly lower detection of Tier 2 viruses (40% vs. PCR 45%). This substitution matters for public health practice for instance, for already recognized zoonotic threats such as Nipah virus, maintaining targeted PCR based diagnostic capacity remains important even as wider Next Generation Sequencing based discovery expands the known viral landscape [50].

Surveillance gaps

Our quantitative results affirm that surveillance uneven distribution creates important gaps in our understanding of global viral ecology and spillover risk [51,52]. The under-sampling of Oceania (249 sequences, 1.5%) and South America (1,371 sequences, 8.0%) is extreme (one-way ANOVA, F (5, 170) = 15.34, p < 0.001).

Ecological and anthropogenic context of virus discovery

The curated dataset's analysis demonstrated the important role that human-modified landscapes play in bat virus surveillance, including dietary guild associations with the virus diversity (Figure 6 A) and high-risk viruses (Figure 6 B), as well as sampling location uneven distribution toward human-modified landscapes (Figure 6 C). Insectivorous bats accounted for most virus records (58.2%), followed by frugivorous bats (29.9%), whereas sanguivorous (90%) and nectarivorous (80%) bats had the highest proportions of high-risk viruses (Tiers 3 and 4) despite their lower overall virus diversity. The geographic distribution of virus detections by bat dietary guild is further detailed in Supplementary Annex 2 D, which shows, for example, that frugivorous bats in Asia and Africa account for the majority of high-risk virus associations outside the America.

A significant number of sampling locations with 70.1% were found in regions with dense populations or recent land-use changes, like agricultural frontiers and fragmented forests. This suggests that human-wildlife interfaces are the focus of an excessive amount of current virus discovery efforts. Also, the majority of source studies (75%) of the viruses categorized as Tier 4 (confirmed zoonotic) identified or hypothesized a known intermediate animal host in the transmission cycle, such as camels for MERS-CoV or pigs for Nipah virus. This finding emphasizes that a complex ecosystem process involving intermediate species at ecological interface zones facilitates spillover for the majority of known bat-borne zoonoses, rather than being a direct bat-to-human event [53].

Sensitivity analysis

Excluding all coronavirus sequences reported after 2020 reduced Coronaviridae prevalence from 42.6% to 28.4%, demonstrating that pandemic driven sequencing significantly influences global diversity patterns.

Drivers of surveillance uneven distribution

Regression analysis indicated that GDP (β = 0.78, p < 0.001, R² = 0.61) and prior outbreak history (β = 0.54, p = 0.003) were significant predictors of sequence deposition per country, whereas bat species richness did not significantly predict sequence deposition (p = 0.18).

Discussion     

Summary of main findings

This scoping review of 17,134 viral genomic sequences associated with bats that were analysed between 2010 and 2024 revealed three main findings. First, surveillance efforts are unevenly distributed where Asia accounts for 39.7% of geolocated sequences while Africa accounts for 29.8%. In contrast, Oceania only had 1.5% and South America had 8.0%, indicating that they have received much less attention. Secondly, three bat families make up 78.6% of all virus records, showing a different research focus and finally, using our evidence-based tiered framework, we found out that 72% of recently discovered viruses have confirmed or high potential zoonotic risk, which falls into Tiers 3 and 4. Our finding that Coronaviridae constitutes 42.6% of bat-associated viral sequences matches a recent meta-analysis of coronavirus surveillance in bats from 1996 to 2019 that also identified Asia and Africa as sampling hotspots.

Geographic and taxonomic patterns

The patterns described above represent differential sampling intensity. When the objective is to assess spillover or outbreak risk, it is rational to prioritize regions and bat species where human-bat contact is frequent and where previous outbreaks have occurred. Sampling strategies that focus on known reservoir taxa (e.g., Rhinolophidae for SARS-related coronaviruses) and high-risk interfaces such as agricultural frontiers, and densely populated areas are appropriate for disease surveillance [54].

However, when the objective is to characterize global viral diversity, understand ecological drivers of viral maintenance, or provide baseline data for early warning systems, this targeted approach leaves substantial gaps. Oceania with 249 sequences representing 1.5% of total and South America with 1,371 sequences, representing 8.0% are undersampled relative to their bat species richness (bias index 0.23 and 0.33, respectively). Bat families such as Emballonuridae, Nycteridae, and Megadermatidae together account for less than 1% of virus records despite representing over 150 species. Our regression analysis shows that GDP (β = 0.78, p < 0.001) and prior outbreak history (β = 0.54, p = 0.003) predict sequence deposition, while bat species richness does not (p = 0.18). This implies that economic capacity and reactive responses to previous emergencies, rather than biodiversity-based prioritization, drive current surveillance. Previous studies have shown that economic capacity and past outbreak experience, rather than biodiversity-based prioritization, are the primary drivers of surveillance effort. The geographic and taxonomic biases identified in this study are consistent with recent reviews that note the disproportionate focus on Rhinolophidae and Pteropodidae as reservoir taxa of concern [55].

Ecological patterns and methodological evolution

The shift from PCR-based to NGS-based methods has expanded known viral diversity but created a functional characterization gap [56]. The efficiency of detection methods varied by virus family, with tissue samples outperforming other sample types for filovirus detection with 87% success rate (Annex 2 E). Feces were the most effective sample type for coronavirus detection (76%). The detection efforts by sample type across all the geographic regions is provided in Annex 1 J.

Our dietary guild analysis showed main ecological patterns where insectivorous bats host the greatest overall viral diversity with 58.2% of the records, while sanguivorous and frugivorous bats showed unbalanced association with high-risk viruses (Tiers 3-4; Annex 1 K; Annex 2 F). Annex 2 F shows the unequal association of frugivorous bats with high-concern zoonotic virus families (filoviruses, henipaviruses) despite lower overall viral diversity.

The temporal transition towards NGS-based detection methods is documented in Supplementary Annex 2 G. PCR based methods declined from 79% (2010) to 5% (2024), while NGS based methods increased from 12% (2010) to 30% (2024). This transition reflects the shift toward agnostic viral discovery but has also created a functional characterization gap for many sequence only viruses. Emerging evidence points to the role of immune system in mediating the unique virus tolerance observed in bats, which may explain their ability to co-exist with high viral loads without disease [57].

The tiered framework for zoonotic risk assessment

We developed an evidence-based four-tier framework to address the inconsistent use of the term "zoonotic" in the literature. When we compared original study labels to our classification, agreement was only 54% for viruses with phylogenetic evidence only (our Tier 2), with original studies frequently overstating risk (e.g., labeling Tier 2 viruses as "zoonotic" without experimental evidence). This heterogeneity (agreement <70% for 12 of 20 virus families) suggests that a standardized framework is needed for reproducible risk communication.

Our framework is not without limitations, as noted in review. First, the 70-90% amino acid identity threshold for Tier 2 is based on published analyses of receptor-binding domain conservation [58], but we acknowledge that functional relevance of sequence identity varies by virus family. Second, the framework represents a snapshot as of November 2024; new experimental evidence can reclassify viruses (e.g., a Tier 2 virus may move to Tier 3 upon demonstration of human cell infection). Third, we did not incorporate additional risk modifiers such as viral replication efficiency, disease severity, or transmission potential. Future iterations could integrate these dimensions. Fourth, we compared our framework to existing systems (World Health Organization (WHO), Centers for Disease Control and Prevention (CDC), Health and Human Services (HHS)); we do not claim novelty of tiered classification but rather offer a synthesis customized to bat-borne viruses with explicit, reproducible thresholds. Despite these limitations, the framework provides a transparent, evidence-graded approach that can be updated as evidence accumulates. We suggest that public databases such as ZOVER could adopt dynamic risk scores ("last reviewed" date + tier) rather than static "zoonotic" labels.

Public health and One Health implications

Recent work has revealed that rapid changes in bat ecology driven anthropogenic activities can trigger spillover events [59]. The reactive nature of current surveillance, demonstrated in temporal spikes following major outbreaks, leaves important blind spots. The 28.4% reduction in Coronaviridae prevalence after excluding post-2020 sequences demonstrate how pandemic-driven sequencing can change perceptions of baseline viral diversity. Therefore, while outbreak responses are important, over reliance on reactive surveillance creates a cycle where funding follows emergencies, undersampled regions remain unexplored, and the next pandemic threat may emerge from a blind spot.

Limitations

We must acknowledge some limitations in this study which include analysis being constrained by the biases inherent in the underlying data where we could only analyze published and database-deposited records. Although iterative refinement of search strategies was done, studies published in non-English regional journals or those not indexed in the databases could have been missed. Also, our tiered classification represents a point-in-time assessment (November 2024) and new experimental evidence could reclassify viruses. Also, the curated dataset (n=244), while stratified and validated, remains a subset of all published studies and may not capture all ecological variability. Last but not least, the exclusion of non-English studies may introduce language bias. Lastly, the 70-90% amino acid identity threshold for Tier 2, while derived from published literature, may not be equally informative across all virus families. Finally, this review is unable to account for negative surveillance findings because our data sources are public sequence databases and the published or preprinted literature, for instance where sampling was conducted but no bat associated virus was detected or sequenced. Therefore, such negative results are rarely deposited or even published, hence the true extent of global surveillance effort and the true degree of geographic unevenness, cannot be fully captured by sequence deposition counts alone.

Conclusion     

This scoping review has revealed that bat-borne virus sequence data does not represent an unbiased number of viral ecological diversity. The available evidence shows differential sampling intensity across geographic regions and bat taxa, driven largely by economic capacity, outbreak experience, and research funding priorities rather than biodiversity-based prioritization. Asia with 39.7% of geolocated sequences and Africa with 29.8% are relatively well sampled, while Oceania with 1.5% and South America accounting 8.0% remain undersampled relative to their bat species richness. Similarly, three bat families, Vespertilionidae, Rhinolophidae, and Pteropodidae account for 78.6% of all virus records, leaving many other families poorly characterized. Applying our evidence-based four-tier framework to recently discovered viruses (2010-2024) revealed that 72% possess confirmed or high-potential zoonotic risk (Tiers 4 and 3, respectively), while 28% have only moderate or low evidence (Tiers 2-1). This finding shows that surveillance that prioritizes high-risk interfaces and known outbreak-associated taxa were oversampled and viruses are perceived as dangerous. The true proportion of zoonotic potential among all bat viruses including those in undisturbed ecosystems and understudied bat families remains unknown. Therefore, addressing current surveillance gaps is a necessary step toward building a reliable early warning system for emerging bat-borne viral threats.

Recommendations

a) Implement stratified sampling by bat dietary guild and roosting ecology in addition to geographic region. b) Develop a funding allocation formula that weights bat species richness, past surveillance effort, and proximity to human-wildlife interfaces. c) Adopt our tiered framework in platforms like ZOVER, with annual evidence updates. d) Support longitudinal, agnostic metagenomic surveys in undisturbed bat colonies per under-sampled continent.

Competing interests     

The authors declare no competing interests.

Authors' contributions     

George Kiprono Kisoi: conceptualization, methodology, formal analysis, writing - original draft, writing - review and editing. Joel Bargul: methodology, validation, writing - review and editing. Jimmy. Nkaiwuatei: data curation, formal analysis, writing - review and editing. Jacob Ngugi Kamau: validation, writing - review and editing. Johnson Kinyua: software, visualization, writing - review and editing. Solomon Langat: supervision, writing - review and editing. Erick Kibichiy Serem: supervision, writing - review and editing. Joel Lutomiah: supervision, resources, writing - review and editing. Fredrick Eyase: supervision, resources, writing - review and editing. All authors have read and approved the final version of this manuscript.

Acknowledgments     

We thank Josphat Tonui, Barbara Bakhita, and Phyllis Waruguru for their expert contribution to the scoping review. We also thank Jorim Oyugi for his contribution to generating a global map that displays the prevalence of bat-borne virus genomic sequences.

Tables and figures     

Table 1: electronic search strategies for bat-borne virus literature (appendices A-H)

Figure 1: PRISMA-ScR flow diagram illustrating the systematic selection process for identifying published articles on bat-borne viruses; numbers indicate records excluded for specific reasons, including duplication and materials being out of the scope of the study objectives (n=1453)

Figure 2: global distribution of bat-associated virus surveillance effort (2010-2024); (A) number of viral genomic sequences deposited in public databases per continent; (B) sequences per bat species, normalized by continental bat species richness

Figure 3: temporal trends in bat virus discovery and surveillance methodology (2010-2024); bars show annual deposition of bat-associated viral sequences in public databases (N = 17,134)

Figure 4: taxonomic distribution of bat virus research by bat family (2010-2024); (A) number of viral sequence records per bat family; (B) records per bat species within each family

Figure 5: evidence-based tiered risk classification of bat-borne viruses and comparison with original literature labels (2010-2024); (A) distribution of recently discovered bat viruses (n = 158 unique species or strains) across the four risk tiers; (B) percentage of viruses labeled zoonotic in the original studies (red bars) compared with our evidence-based classification (blue bars)

Figure 6: ecological patterns in bat-virus associations from the curated dataset (2010-2024); (A) percentage of virus records by bat dietary guild; (B) percentage of high-risk viruses (Tiers 3 and 4) within each dietary guild; (C) percentage of sampling locations in human-modified versus undisturbed or natural habitats; data are from the curated subset (n = 244 bat-virus association records)

Annexes     

Annex 1: supplementary tables (PDF - 511KB)

Annex 2: supplementary figures (PDF - 1008KB)

References