The Sparsely Sampled Ubiquity of Global Fungal Endophytes

The Sparsely Sampled Ubiquity of Global Fungal Endophytes | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Biological Sciences - Article The Sparsely Sampled Ubiquity of Global Fungal Endophytes Beatrice Bock, Nicholas McKay, Nancy Johnson, Catherine Gehring This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8244775/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract The paradigm that ‘all plants harbor fungal endophytes’ is a widely accepted axiom in plant ecology, yet this claim has never been tested. Here, we use a machine learning pipeline to screen 21,891 research abstracts spanning nearly a century. We developed a high-sensitivity screening tool to find and manually validate all potential claims of endophyte absence. This comprehensive validation revealed no evidence of a verifiable endophyte-free plant taxon within the entire scientific literature surveyed. This finding confirms that endophyte ubiquity is a robust pattern wherever plants have been studied. However, we show that this “ubiquitous” paradigm is a generalization from an extraordinarily biased sample: analysis of the 19,119 relevant primary studies reveals that only 0.8% of described plant species have been examined for fungal endophytes, with 77% of this research concentrated in the Global North. Our analysis reframes a foundational paradigm, revealing that our understanding of a “global” symbiosis is based on a tiny fraction of biodiversity. This sampling bias represents a critical bottleneck in biodiversity research, obscuring the discovery of novel biotechnologies and climate-resilient symbioses in the world’s most threatened ecosystems. Biological sciences/Ecology/Macroecology Biological sciences/Ecology/Conservation biology Biological sciences/Ecology/Biodiversity Biological sciences/Ecology/Biogeography Biological sciences/Ecology/Microbial ecology Figures Figure 1 Figure 2 Figure 3 Introduction Fungal endophytes represent one of the largest and most chemically diverse microbial reservoirs on Earth 1 . These symbionts, distinguished from mycorrhizal fungi by their ability to inhabit broad plant tissues without forming specific root interfaces 2 and without causing disease 3 , are not merely passive passengers; they actively modulate plant stress tolerance, drive ecosystem productivity, and produce a vast array of bioactive compounds with pharmaceutical and agricultural applications 4,5 . Underpinning this functional potential is the foundational paradigm that this symbiosis is ubiquitous. Influential book chapters (e.g., 6 ) and the introductions of highly cited research articles (e.g., 7 ) state this claim explicitly. This idea has become a central, defining assumption in ecology, which is restated in contemporary reviews (e.g., 2 ) noting they are “found in nearly all plants,” and underpins research on endophyte function, diversity, and transmission. However, this assumption of ubiquity remains unvalidated, as a comprehensive, systematic evaluation of fungal endophyte distribution across plant taxa and regions has never been conducted. Previous reviews have been qualitative or focused on specific lineages (e.g., 8,9 ), leaving the global pattern untested. Crucially, testing this paradigm requires distinguishing between a “study-level absence” (a failure to detect endophytes in a specific sample or experiment) and a “taxon-level absence” (a plant taxon that always lacks endophytes). While study-level absences are expected due to detection limits and differing methodological approaches, a confirmed taxon-level absence would fundamentally challenge the ubiquity paradigm. Conversely, confirming ubiquity while revealing major sampling gaps would expose a critical bottleneck in our ability to bioprospect this global reservoir for novel functions. Here, we leverage a novel machine learning pipeline to conduct the first large-scale, quantitative test of endophyte ubiquity. We screen 21,891 abstracts to systematically test the claim of ubiquity and analyze the taxonomic and geographic patterns within the 19,119 identified primary research articles. Our analysis confirms endophyte ubiquity yet reveals it is founded on a dataset that ignores the vast majority of plant biodiversity. Results Our analysis of 21,891 abstracts confirms that the literature is overwhelmingly a record of endophyte presence. Given that ‘true absence’ is an exceptionally rare event, we designed a high-sensitivity screening pipeline to find and manually validate all potential absence candidates. To rigorously test for absence, we applied a two-stage machine learning pipeline to the full corpus of 21,891 abstracts. We first filtered for relevance, removing 11% of the corpus consisting of secondary literature (e.g., reviews) and non-fungal studies, to isolate a final analytical dataset of 19,119 primary research articles. To identify rare negative results within this relevant corpus, we employed a high-sensitivity ensemble classifier 10,11 designed to prioritize potential absence claims. This screening tool flagged 89 candidate abstracts. Combined with an independent keyword search (see Methods), this yielded a total of 102 candidates (0.5% of the relevant literature) for manual review. We manually validated all 102 of these cases. This comprehensive curation revealed that none represented a true, taxon-level absence. While we identified individual studies that failed to find endophytes (study-level absences), further investigation revealed that other studies published on the same plant taxon reported presence. Consequently, our analysis found no verifiable examples of an endophyte-free plant taxon in the entire 21,891-abstract dataset. A Biased View of Ubiquity Despite this confirmed ubiquity, our analysis reveals that this conclusion rests on a tiny and deeply biased sample of plant biodiversity. We find that only 0.8% (3,226 of 390,101 species) of described plant species have been examined for endophytes within the relevant corpus of 19,119 abstracts (Figure 1). This sparsity is consistent across all taxonomic ranks: only 6.7% (1,506 of 22,441) of described plant genera (Supp. Fig. 1) and only 28.9% (343 of 1,185) of described plant families have been studied for endophytes (Supp. Fig. 2). The corresponding study of the fungal lineages studied in the dataset only reached 3.5% of described fungal species (5,064 of 143,957 species) (Supp. Fig. 3). This bias systematically overlooks critical evolutionary lineages. For example, Charophyta (stoneworts), the closest algal relatives to land plants, have no coverage (0 of 2,012 species) in our dataset. This is a critical knowledge gap, as it leaves the aquatic origins of plant-fungal symbiosis almost completely untested, which is a relationship thought to predate the colonization of land by millions of years 12,13 . Similarly, Bryophyta (mosses), representing some of the earliest land plants, have only 0.2% coverage (21 of 10,508 species). This gap obscures our understanding of how these symbioses evolved after plants first colonized land. Even within the relatively well-studied Tracheophyta (vascular plants), which encompasses familiar groups like ferns, conifers, and flowering plants, coverage remains exceptionally low at 0.9% (3,150 of 363,445 species). This analysis demonstrates that our knowledge of endophyte ubiquity is largely derived from a thin sliver of plant diversity. This taxonomic bias is mirrored by an extreme geographic bias (Figure 2; Supp. Fig. 4). Research is concentrated in high-income countries (the Global North), which account for 77% of all geographically coded studies 14 . This concentration is driven almost entirely by just two regions: North America and Europe alone account for 70.3% of all studies. Conversely, entire nations with unique biodiversity, such as Tonga, have zero studies detected in our dataset. A temporal analysis shows that these geographic and taxonomic biases are not historical artifacts; they have remained largely consistent over the past three decades, even as molecular methods have become standard (used in 36.8% of all studies; Figure 3; Supp. Fig. 5). Finally, analysis of plant tissues reported in the abstracts shows that research has covered a wide range, with roots, leaves, seeds, stems, shoots, and fruits being the most frequently studied structures (Supp. Fig. 6). This demonstrates thorough sampling across tissue types within the limited taxonomic breadth. Discussion Our analysis provides the first large-scale, data-driven confirmation of the fungal endophyte ubiquity paradigm. By demonstrating that there is no evidence for a single endophyte-free plant taxon in the literature, we confirm that ubiquity is a robust pattern wherever plants have been studied. However, the central finding of this paper is the stark contradiction between this observed ubiquity and the severe sampling biases that define the field. The 0.8% of plant species and 77% concentration in the Global North are not minor gaps; they are systematic biases that fundamentally distort our view of global fungal ecology. Our “ubiquitous” paradigm is a generalization from a non-random sample shaped largely by research infrastructure and funding availability rather than biological priority. Our data strongly suggest that the true biodiversity of fungal endophytes is severely underestimated, a conclusion reinforced by both biological and methodological constraints. First, standard definitions of ‘endophyte’ exclude obligate symbionts such as Glomeromycota (arbuscular mycorrhizal fungi); thus, our dataset represents a conservative lower bound of plant-associated fungal diversity. Second, despite analyzing 21,891 abstracts, we found evidence for only 3.5% of described fungal species (5,064 of 143,957) and 16.5% of fungal genera (Supp. Fig. 3). Given that we have sampled only 0.8% of global plant diversity, the ‘missing’ fungal biodiversity likely resides within the 99.2% of known plant species that remain unexplored. This sampling gap has fundamental implications for estimates of global fungal richness, which rely heavily on extrapolations of plant-to-fungus ratios 15 . Our data suggest that current global estimates, derived largely from the well-studied assemblages of the Global North, cannot be reliably extrapolated to the distinct environmental and evolutionary contexts of the tropical and arid hosts that dominate the global flora. Current estimates of global fungal richness may therefore require significant upward revision as we expand sampling beyond the few dominant plant hosts currently represented in the literature. From “Study-Level Negatives” to “Taxon-Level Ubiquity” A key challenge in testing ubiquity is the interpretation of “absence” reports. Our screening pipeline flagged 102 abstracts that indicated absence. Manual validation clarified that these were either irrelevant (40), simple false positives (i.e., abstracts incorrectly classified as ‘Absence’ by the model; 46), or reports of mixed results (15) where endophytes were found in some tissues but not others or in some plant individuals but not others. We identified only a single genuine “study-level absence” in the literature: a study reporting no fungal endophytes in Phragmites australis 16 . However, this absence is contradicted by numerous other studies that confirm diverse fungal endophyte communities in the same host species(e.g., 17–19 ). This specific case perfectly illustrates our core distinction: while individual studies may fail to detect endophytes (“study-level negatives”), “taxon-level ubiquity” holds true. We found no evidence of a plant taxon that is consistently free of fungal endophytes. Limitations of Scope Our analysis relies on metadata and claims extracted from scientific abstracts. While this approach enables a scale of analysis impossible with manual full-text review, it inherently produces a conservative dataset, as abstracts may not explicitly list every host species or collection site. However, given that coverage stands at only 0.8% of plant species, even a substantial increase in species recovery from full texts would not alter the fundamental conclusion: the vast majority of global plant biodiversity remains completely uninvestigated. A critical factor influencing the published record is publication bias, often termed the 'file drawer problem' 20 . In the context of the ubiquity paradigm, this bias is likely circular: failure to detect endophytes is often attributed to methodological error rather than biological absence, discouraging the publication of negative results. Thus, our ‘zero verifiable absences’ finding reflects the consensus of the published record, which is itself shaped by the assumption of ubiquity. Finally, our restriction to English-language abstracts imposes a linguistic filter on the map of knowledge. Consequently, the ‘zero’ counts observed in many nations in the Global South (e.g., across substantial portions of Africa) likely reflect a compound failure: a genuine scarcity of funded biodiversity monitoring, the exclusion of non-English records within indexed databases, and a systemic indexing bias that omits regional journals entirely. Conclusion This concentration of research has profound practical consequences. By prioritizing the Global North, we are effectively ignoring the tropical and arid biodiversity hotspots of the Global South 21 , regions where plants face some of the most extreme environmental pressures 22 . This geographic bias is particularly critical given the contrasting biogeography of fungal guilds. While the diversity of key soil guilds, particularly ectomycorrhizal fungi (definitionally excluded from being endophytes), peaks in temperate and boreal zones 23 , foliar endophytes appear to follow the canonical latitudinal diversity gradient, reaching peak richness in the tropics 7 . By concentrating 77% of research in the Global North, the field has inadvertently aligned its sampling effort with the diversity centers of mycorrhizae, while systematically ignoring the hyperdiverse centers of endophytic fungi. This reframes the research bias from a simple academic gap to a significant bottleneck in global bioprospecting 5,24 . The cost of this omission is illustrated by recent successes in under-sampled, extreme environments; for instance, endophytes from the Antarctic angiosperm Colobanthus quitensis have yielded unique metabolic pathways with potential applications in neurodegenerative disease treatment 25 . Such discoveries confirm that the 99% of plant hosts currently excluded from our limited, sampling biased toward the Global North represent a vast, untapped reservoir of functional biology. Beyond quantifying this bias, our analysis provides a corrective tool: a data-driven roadmap of the plant taxa and geographic regions that define the ‘missing’ majority of the literature. Provided in the Supplementary Information, this resource identifies priority targets, such as the phyla Charophyta and Bryophyta and nations like Tonga, where directed sampling is most likely to maximize novelty. Having established the global ubiquity of fungal endophytes, the field must now pivot from verifying their presence to exploring their function. By redirecting research toward these neglected lineages and regions, we can unlock the next generation of climate-resilient symbioses and biotechnologies hidden within the world’s most threatened ecosystems. Methods Our methodology followed a three-phase computational workflow, beginning with the full dataset of 21,891 abstracts. Phase 1: High-Sensitivity Screening. We developed a two-stage machine learning pipeline to first isolate relevant primary research (n=19,447) from the full dataset (n=21,891) and then classify these abstracts for potential absence claims. Phase 2: Human Validation. We manually reviewed all abstracts flagged as potential absences (n=102) to verify the ‘zero absence’ finding. Phase 3: Extraction & Mapping. We applied a validated extraction pipeline to the relevant corpus (n=19,447). To ensure our bias analysis focused strictly on endophytes, we excluded abstracts focused solely on mycorrhizal fungi, resulting in a final analytical dataset of 19,199 abstracts for taxonomic and geographic mapping. 1. Machine Learning Classification as a High-Sensitivity Screen For phase 1 of our workflow, we developed a two-stage machine learning pipeline. The first stage screened for relevance, resulting in the 19,447 abstract dataset. The second stage classified relevant abstracts as either indicating presence or absence of fungal endophytes in the studied plants. For this classification, we faced an extreme imbalanced class problem: our manually curated 876-abstract training dataset contained 346 ‘Presence’ examples but only a single ‘Absence’ example (0.11%). To address this scarcity, we employed data augmentation techniques, generating synthetic abstracts representing plausible absence claims using large language models (see Supplementary Methods). These synthetic data were used solely to define the linguistic boundaries during training and were not part of the analyzed dataset. To design a high-sensitivity screening tool, we used a weighted ensemble of two machine learning models: a Support Vector Machine (a model that finds the optimal boundary separating two groups 26–28 ) and a regularized logistic regression (a model that automatically selects the most important predictive words 29–31 ). As documented in our model testing, we intentionally weighted the model to prioritize absence detection, creating a screen that would flag all potential candidates for manual review. 2. Pipeline Validation: Absence and Manual Review We implemented a rigorous, multi-step process to validate all potential “Absence” classifications. First, our Machine Learning (ML) screen flagged 89 candidates. Second, an independent rule-based string detection algorithm flagged 12 additional unique candidates. Combined with the inclusion of our single positive control from the training set, this created a final set of 102 abstracts for manual validation. We manually reviewed all of these 102 abstracts. The results, detailed in the Supplementary Methods, confirmed that zero represented a true, taxon-level absence. 3. Data Extraction Pipeline and Validation We built a modular data extraction pipeline to parse taxonomic, geographic, and methodological information. The robustness of every component, from species name extraction to geographic coordinate parsing, was verified using a comprehensive, automated test suite. This suite evaluated core functionality, accuracy, synonym resolution, and error handling, ensuring high fidelity in the extracted data. After taxonomic information was attached to taxon names in the abstracts, any taxa listed as mycorrhizal in FungalTraits 32 and any taxa with the phylum Glomeromycota were removed. Although mycorrhizal fungi inhabit plant tissues, they represent a distinct, well-characterized functional guild separate from the general “endophyte” paradigm tested here 2 ; therefore, abstracts focusing exclusively on mycorrhizal fungi were excluded to avoid confounding the results on endophytic fungi. 4. Data Aggregation Taxonomic names were standardized, and synonyms were resolved against the Global Biodiversity Information Facility (GBIF) backbone 33 to ensure accurate aggregation of species counts. We used the Paleobiology Database 34 to exclude extinct taxa from the GBIF backbone. Geographic information was aggregated at the country and region level. For regional analysis, countries were classified as ‘Global North’ or ‘Global South’ based on the World Bank’s income group classifications (e.g., ‘High income’ vs. ‘Low/Middle income’) 14 . Declarations Author Contributions B.B. conceived the study, designed the methodology, performed the data analysis and machine learning, and wrote the manuscript. N.P.M., N.C.J. and C.A.G. contributed to the final manuscript text. All authors read and approved the final manuscript. Competing Interests The authors declare no competing interests. Funding Statement This work was supported by National Science Foundation (NSF) MacroSystems grant DEB-1340852 (C.A.G.), US Department of Energy program in Systems Biology Research to Advance Sustainable Bioenergy Crop Development (DE-FOA-0002214)(N.C.J.), the Lucking Family Professorship (C.A.G.), the ARCS Foundation (B.M.B.), the Presidential Fellowship Program at NAU (B.M.B.), the Arizona Mushroom Society (B.M.B.), the Support for Graduate Students program at NAU (B.M.B.), and the American Association of University Women’s American Dissertation Fellowship (B.M.B.). Data and Code Availability The full code for the data analysis pipeline, including all R scripts used for machine learning, validation, and figure generation, is available at DOI: 10.5281/zenodo.17770500 35 . The final abstract dataset and model outputs generated during this study are available at DOI: 10.5281/zenodo.17770522 36 . References Kusari, S., Hertweck, C. & Spiteller, M. Chemical Ecology of Endophytic Fungi: Origins of Secondary Metabolites. Chem. Biol. 19 , 792–798 (2012). Cosner, J. et al. Fungal endophytes. Curr. Biol. 35 , R904–R910 (2025). Petrini, O. Fungal Endophytes of Tree Leaves. in Microbial Ecology of Leaves (eds Andrews, J. H. & Hirano, S. S.) 179–197 (Springer New York, New York, NY, 1991). doi:10.1007/978-1-4612-3168-4_9. Rodriguez, R. J., White Jr, J. F., Arnold, A. E. & Redman, R. S. Fungal endophytes: diversity and functional roles. New Phytol. 182 , 314–330 (2009). Strobel, G. A. Endophytes as sources of bioactive products. Microbes Infect. 5 , 535–544 (2003). Stone, J. K., Bacon, C. W. & White, F. An Overview of Endophytic Microbes: Endophytism Defined. in Microbial Endophytes 1–28 (CRC Press, 2000). Arnold, A. & Lutzoni, F. Diversity and host range of foliar fungal endophytes: Are tropical leaves biodiversity hotspots? Ecology 88 , 541–549 (2007). Schulz, B. & Boyle, C. The endophytic continuum. Mycol. Res. 109 , 661–686 (2005). U’Ren, J. M., Lutzoni, F., Miadlikowska, J., Laetsch, A. D. & Arnold, A. E. Host and geographic structure of endophytic and endolichenic fungi at a continental scale. Am. J. Bot. 99 , 898–914 (2012). Sagi, O. & Rokach, L. Ensemble learning: A survey. WIREs Data Min. Knowl. Discov. 8 , e1249 (2018). Shah, M. et al. Theoretical Evaluation of Ensemble Machine Learning Techniques. in 2023 5th International Conference on Smart Systems and Inventive Technology (ICSSIT) 829–837 (IEEE, Tirunelveli, India, 2023). doi:10.1109/ICSSIT55814.2023.10061139. Martin, F. M., Uroz, S. & Barker, D. G. Ancestral alliances: Plant mutualistic symbioses with fungi and bacteria. Science 356 , eaad4501 (2017). Strullu‐Derrien, C., Selosse, M., Kenrick, P. & Martin, F. M. The origin and evolution of mycorrhizal symbioses: from palaeomycology to phylogenomics. New Phytol. 220 , 1012–1030 (2018). World Bank. World Bank Country and Lending Groups. (2025). Hawksworth, D. L. & Lücking, R. Fungal Diversity Revisited: 2.2 to 3.8 Million Species. Microbiol. Spectr. 5 , 5.4.10 (2017). Lambert, A. & Casagrande, R. No evidence of fungal endophytes in native and exotic Phragmites australis. Northeast. Nat. 13 , 561–568 (2006). Wirsel, S. G. R., Leibinger, W., Ernst, M. & Mendgen, K. Genetic diversity of fungi closely associated with common reed. New Phytol. 149 , 589–598 (2001). Angelini, P. et al. The endophytic fungal communities associated with the leaves and roots of the common reed (Phragmites australis) in Lake Trasimeno (Perugia, Italy) in declining and healthy stands. Fungal Ecol. 5 , 683–693 (2012). Clay, K., Shearin, Z., Bourke, K., Bickford, W. & Kowalski, K. Diversity of fungal endophytes in non-native Phragmites australis in the Great Lakes. Biol. Invasions 18 , 2703–2716 (2016). Rosenthal, R. The file drawer problem and tolerance for null results. Psychol. Bull. 86 , 638–641 (1979). Myers, N., Mittermeier, R. A., Mittermeier, C. G., Da Fonseca, G. A. B. & Kent, J. Biodiversity hotspots for conservation priorities. Nature 403 , 853–858 (2000). Seneviratne, S. I. & Zhang, X. Weather and Climate Extreme Events in a Changing Climate. in Climate Change 2021 – The Physical Science Basis: Working Group I Contribution to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change (Cambridge University Press, 2023). doi:10.1017/9781009157896. Tedersoo, L. et al. Global diversity and geography of soil fungi. Science 346 , 1256688 (2014). Newman, D. J. & Cragg, G. M. Natural Products As Sources of New Drugs over the 30 Years from 1981 to 2010. J. Nat. Prod. 75 , 311–335 (2012). Bertini, L. et al. Biodiversity and Bioprospecting of Fungal Endophytes from the Antarctic Plant Colobanthus quitensis. J. Fungi 8 , (2022). Joachims, T. A statistical learning learning model of text classification for support vector machines. in Proceedings of the 24th annual international ACM SIGIR conference on Research and development in information retrieval 128–136 (ACM, New Orleans Louisiana USA, 2001). doi:10.1145/383952.383974. Drake, J. M., Randin, C. & Guisan, A. Modelling ecological niches with support vector machines. J. Appl. Ecol. 43 , 424–432 (2006). Mammone, A., Turchi, M. & Cristianini, N. Support vector machines. WIREs Comput. Stat. 1 , 283–289 (2009). Genkin, A., Lewis, D. D. & Madigan, D. Large-Scale Bayesian Logistic Regression for Text Categorization. Technometrics 49 , 291–304 (2007). Salehi, F., Abbasi, E. & Hassibi, B. The Impact of Regularization on High-dimensional Logistic Regression. Preprint at https://doi.org/10.48550/arXiv.1906.03761 (2019). Ichsan, A., Riyadi, S. & Pardede, D. Analysis of Logistic Regression Regularization in Wild Elephant Classification with VGG-16 Feature Extraction. J. Comput. Netw. Archit. High Perform. Comput. 6 , 783–793 (2024). Põlme, S. et al. FungalTraits: a user-friendly traits database of fungi and fungus-like stramenopiles. Fungal Divers. 105 , 1–16 (2020). Registry-Migration.Gbif.Org. GBIF Backbone Taxonomy. GBIF Secretariat https://doi.org/10.15468/39OMEI (2023). Uhen, M. D. et al. Paleobiology Database User Guide Version 1.0. PaleoBios 40 , (2023). Bock, B. M. beabock/Endo-Review: endo-review: a ML pipeline to review fungal endophyte ubiquity. Zenodo https://doi.org/10.5281/ZENODO.17770500 (2025). Bock, B. M. Endo-Review: Data, Models, and Results. Zenodo https://doi.org/10.5281/ZENODO.17770522 (2025). Chawla, N. V., Bowyer, K. W., Hall, L. O. & Kegelmeyer, W. P. SMOTE: Synthetic Minority Over-sampling Technique. J. Artif. Intell. Res. 16 , 321–357 (2002). Kuhn, M. caret: Classification and Regression Training. 7.0–1 https://doi.org/10.32614/CRAN.package.caret (2007). Raasveldt, M. & Mühleisen, H. DuckDB: an Embeddable Analytical Database. in Proceedings of the 2019 International Conference on Management of Data 1981–1984 (ACM, Amsterdam Netherlands, 2019). doi:10.1145/3299869.3320212. R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing (2023). Wickham, H. Ggplot2: Elegant Graphics for Data Analysis . (Springer international publishing, Cham, 2016). Additional Declarations There is NO Competing Interest. Supplementary Files SupplementaryMethods.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8244775","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Biological Sciences - Article","associatedPublications":[],"authors":[{"id":553048795,"identity":"622db419-6da7-413e-b71d-61c885e8b96a","order_by":0,"name":"Beatrice Bock","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAsUlEQVRIiWNgGAWjYBACAwYexgcSFTYQHg+RWpgNLM6kkaaFTaCy5TAJWszFzh5juNlwPo9/RgLjg7dtRGixnJ2X9nDmjtvFEjcSmA3nEqPF4HaOubHkmduJDTcS2KR5idRiJv237Vzi/BsJ7L+J1iIh2XYgcQPQFmaitAD9kmwgcSY5ceOZh82Sc84RocVcOvcgMCrtEucdTz744U0ZEVqQAGMDaepHwSgYBaNgFOAGAP5mOsJOUpg4AAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0003-2240-9360","institution":"Northern Arizona University","correspondingAuthor":true,"prefix":"","firstName":"Beatrice","middleName":"","lastName":"Bock","suffix":""},{"id":553048796,"identity":"8efde1b1-6beb-4fc6-8f90-c1d09fc78c7e","order_by":1,"name":"Nicholas McKay","email":"","orcid":"","institution":"School of Earth and Sustainability, Northern Arizona University","correspondingAuthor":false,"prefix":"","firstName":"Nicholas","middleName":"","lastName":"McKay","suffix":""},{"id":553048797,"identity":"09c161f4-fad8-4bb1-a9d1-d66448cb7de4","order_by":2,"name":"Nancy Johnson","email":"","orcid":"https://orcid.org/0000-0001-7077-6232","institution":"Northern Arizona University","correspondingAuthor":false,"prefix":"","firstName":"Nancy","middleName":"","lastName":"Johnson","suffix":""},{"id":553048798,"identity":"498a92c6-0cf5-4ad9-906f-a4e224edb4c2","order_by":3,"name":"Catherine Gehring","email":"","orcid":"https://orcid.org/0000-0002-9393-9556","institution":"Northern Arizona University","correspondingAuthor":false,"prefix":"","firstName":"Catherine","middleName":"","lastName":"Gehring","suffix":""}],"badges":[],"createdAt":"2025-11-30 23:35:14","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8244775/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8244775/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":97253821,"identity":"e129ff1f-fcab-487d-8d29-e69a451052e8","added_by":"auto","created_at":"2025-12-02 13:25:12","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":262066,"visible":true,"origin":"","legend":"\u003cp\u003ePlant species representation by phylum in the analyzed literature. Bars show the percentage of described species within each phylum present in the 19,119 relevant primary studies. Overall coverage = 0.8% (3,226 of 390,101 described plant species).\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8244775/v1/320945c8eb37c041142b4d1c.png"},{"id":97253598,"identity":"460f88c4-0d51-468b-b836-1316c4dc6223","added_by":"auto","created_at":"2025-12-02 13:25:04","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":386041,"visible":true,"origin":"","legend":"\u003cp\u003eGlobal geographic bias in fungal endophyte research. The heatmap displays the frequency of studies per country (log scale), where darker colors indicate higher research intensity and light gray indicates zero studies detected in the analyzed corpus. The distribution reveals a stark concentration of research in the Global North (e.g., North America and Europe), while vast regions of the Global South, including large portions of Africa and Southeast Asia, remain virtually invisible in the indexed scientific literature.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8244775/v1/ad9eb5f5deef02a416bfb940.png"},{"id":97253642,"identity":"a8218a1c-5a10-4487-aa7b-7df708cc55cf","added_by":"auto","created_at":"2025-12-02 13:25:06","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":111926,"visible":true,"origin":"","legend":"\u003cp\u003eTemporal trends in methods reported for detecting fungal endophytes across the 21,891-abstract corpus. Molecular approaches were negligible before the 1990s and rose sharply after 2000; overall, 36.8% of studies in the corpus report molecular methods.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-8244775/v1/d14e4ef0b32c14c3e7ecd006.png"},{"id":97664496,"identity":"fb124fb0-b585-4a34-923d-0281e3f0032c","added_by":"auto","created_at":"2025-12-08 09:05:53","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1220154,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8244775/v1/e26e0237-a773-4835-9e05-6d9b4bf82587.pdf"},{"id":97253637,"identity":"24ec3413-63ea-4c61-9668-5745f993cd42","added_by":"auto","created_at":"2025-12-02 13:25:06","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1446524,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMethods.docx","url":"https://assets-eu.researchsquare.com/files/rs-8244775/v1/8041d52e162bb00071ec6026.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"The Sparsely Sampled Ubiquity of Global Fungal Endophytes","fulltext":[{"header":"Introduction","content":"\u003cp\u003eFungal endophytes represent one of the largest and most chemically diverse microbial reservoirs on Earth\u003csup\u003e1\u003c/sup\u003e. These symbionts, distinguished from mycorrhizal fungi by their ability to inhabit broad plant tissues without forming specific root interfaces\u003csup\u003e2\u003c/sup\u003e and without causing disease\u003csup\u003e3\u003c/sup\u003e, are not merely passive passengers; they actively modulate plant stress tolerance, drive ecosystem productivity, and produce a vast array of bioactive compounds with pharmaceutical and agricultural applications\u003csup\u003e4,5\u003c/sup\u003e. Underpinning this functional potential is the foundational paradigm that this symbiosis is ubiquitous. Influential book chapters (e.g.,\u003csup\u003e6\u003c/sup\u003e) and the introductions of highly cited research articles (e.g.,\u003csup\u003e7\u003c/sup\u003e) state this claim explicitly. This idea has become a central, defining assumption in ecology, which is restated in contemporary reviews (e.g.,\u003csup\u003e2\u003c/sup\u003e) noting they are \u0026ldquo;found in nearly all plants,\u0026rdquo; and underpins research on endophyte function, diversity, and transmission.\u003c/p\u003e\n\u003cp\u003eHowever, this assumption of ubiquity remains unvalidated, as a comprehensive, systematic evaluation of fungal endophyte distribution across plant taxa and regions has never been conducted. Previous reviews have been qualitative or focused on specific lineages (e.g.,\u003csup\u003e8,9\u003c/sup\u003e), leaving the global pattern untested. Crucially, testing this paradigm requires distinguishing between a \u0026ldquo;study-level absence\u0026rdquo; (a failure to detect endophytes in a specific sample or experiment) and a \u0026ldquo;taxon-level absence\u0026rdquo; (a plant taxon that always lacks endophytes). While study-level absences are expected due to detection limits and differing methodological approaches, a confirmed taxon-level absence would fundamentally challenge the ubiquity paradigm. Conversely, confirming ubiquity while revealing major sampling gaps would expose a critical bottleneck in our ability to bioprospect this global reservoir for novel functions.\u003c/p\u003e\n\u003cp\u003eHere, we leverage a novel machine learning pipeline to conduct the first large-scale, quantitative test of endophyte ubiquity. We screen 21,891 abstracts to systematically test the claim of ubiquity and analyze the taxonomic and geographic patterns within the 19,119 identified primary research articles. Our analysis confirms endophyte ubiquity yet reveals it is founded on a dataset that ignores the vast majority of plant biodiversity.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eOur analysis of 21,891 abstracts confirms that the literature is overwhelmingly a record of endophyte presence. Given that \u0026lsquo;true absence\u0026rsquo; is an exceptionally rare event, we designed a high-sensitivity screening pipeline to find and manually validate all potential absence candidates.\u003c/p\u003e\n\u003cp\u003eTo rigorously test for absence, we applied a two-stage machine learning pipeline to the full corpus of 21,891 abstracts. We first filtered for relevance, removing 11% of the corpus consisting of secondary literature (e.g., reviews) and non-fungal studies, to isolate a final analytical dataset of 19,119 primary research articles. To identify rare negative results within this relevant corpus, we employed a high-sensitivity ensemble classifier\u003csup\u003e10,11\u003c/sup\u003e designed to prioritize potential absence claims.\u003c/p\u003e\n\u003cp\u003eThis screening tool flagged 89 candidate abstracts. Combined with an independent keyword search (see Methods), this yielded a total of 102 candidates (0.5% of the relevant literature) for manual review. We manually validated all 102 of these cases. This comprehensive curation revealed that none represented a true, taxon-level absence. While we identified individual studies that failed to find endophytes (study-level absences), further investigation revealed that other studies published on the same plant taxon reported presence. Consequently, our analysis found no verifiable examples of an endophyte-free plant taxon in the entire 21,891-abstract dataset.\u003c/p\u003e\n\u003ch3\u003eA Biased View of Ubiquity\u003c/h3\u003e\n\u003cp\u003eDespite this confirmed ubiquity, our analysis reveals that this conclusion rests on a tiny and deeply biased sample of plant biodiversity. We find that only 0.8% (3,226 of 390,101 species) of described plant species have been examined for endophytes within the relevant corpus of 19,119 abstracts (Figure 1). This sparsity is consistent across all taxonomic ranks: only 6.7% (1,506 of 22,441) of described plant genera (Supp. Fig. 1) and only 28.9% (343 of 1,185) of described plant families have been studied for endophytes (Supp. Fig. 2). The corresponding study of the fungal lineages studied in the dataset only reached 3.5% of described fungal species (5,064 of 143,957 species) (Supp. Fig. 3).\u003c/p\u003e\n\u003cp\u003eThis bias systematically overlooks critical evolutionary lineages. For example, Charophyta (stoneworts), the closest algal relatives to land plants, have no coverage (0 of 2,012 species) in our dataset. This is a critical knowledge gap, as it leaves the aquatic origins of plant-fungal symbiosis almost completely untested, which is a relationship thought to predate the colonization of land by millions of years\u003csup\u003e12,13\u003c/sup\u003e. Similarly, Bryophyta (mosses), representing some of the earliest land plants, have only 0.2% coverage (21 of 10,508 species). This gap obscures our understanding of how these symbioses evolved after plants first colonized land.\u003c/p\u003e\n\u003cp\u003eEven within the relatively well-studied Tracheophyta (vascular plants), which encompasses familiar groups like ferns, conifers, and flowering plants, coverage remains exceptionally low at 0.9% (3,150 of 363,445 species). This analysis demonstrates that our knowledge of endophyte ubiquity is largely derived from a thin sliver of plant diversity.\u003c/p\u003e\n\u003cp\u003eThis taxonomic bias is mirrored by an extreme geographic bias (Figure 2; Supp. Fig. 4). Research is concentrated in high-income countries (the Global North), which account for 77% of all geographically coded studies\u003csup\u003e14\u003c/sup\u003e. This concentration is driven almost entirely by just two regions: North America and Europe alone account for 70.3% of all studies. Conversely, entire nations with unique biodiversity, such as Tonga, have zero studies detected in our dataset. A temporal analysis shows that these geographic and taxonomic biases are not historical artifacts; they have remained largely consistent over the past three decades, even as molecular methods have become standard (used in 36.8% of all studies; Figure 3; Supp. Fig. 5).\u003c/p\u003e\n\u003cp\u003eFinally, analysis of plant tissues reported in the abstracts shows that research has covered a wide range, with roots, leaves, seeds, stems, shoots, and fruits being the most frequently studied structures (Supp. Fig. 6). This demonstrates thorough sampling across tissue types within the limited taxonomic breadth.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eOur analysis provides the first large-scale, data-driven confirmation of the fungal endophyte ubiquity paradigm. By demonstrating that there is no evidence for a single endophyte-free plant taxon in the literature, we confirm that ubiquity is a robust pattern wherever plants have been studied.\u003c/p\u003e\n\u003cp\u003eHowever, the central finding of this paper is the stark contradiction between this observed ubiquity and the severe sampling biases that define the field. The 0.8% of plant species and 77% concentration in the Global North are not minor gaps; they are systematic biases that fundamentally distort our view of global fungal ecology. Our \u0026ldquo;ubiquitous\u0026rdquo; paradigm is a generalization from a non-random sample shaped largely by research infrastructure and funding availability rather than biological priority.\u003c/p\u003e\n\u003cp\u003eOur data strongly suggest that the true biodiversity of fungal endophytes is severely underestimated, a conclusion reinforced by both biological and methodological constraints. First, standard definitions of \u0026lsquo;endophyte\u0026rsquo; exclude obligate symbionts such as Glomeromycota (arbuscular mycorrhizal fungi); thus, our dataset represents a conservative lower bound of plant-associated fungal diversity. Second, despite analyzing 21,891 abstracts, we found evidence for only 3.5% of described fungal species (5,064 of 143,957) and 16.5% of fungal genera (Supp. Fig. 3).\u003c/p\u003e\n\u003cp\u003eGiven that we have sampled only 0.8% of global plant diversity, the \u0026lsquo;missing\u0026rsquo; fungal biodiversity likely resides within the 99.2% of known plant species that remain unexplored. This sampling gap has fundamental implications for estimates of global fungal richness, which rely heavily on extrapolations of plant-to-fungus ratios\u003csup\u003e15\u003c/sup\u003e. Our data suggest that current global estimates, derived largely from the well-studied assemblages of the Global North, cannot be reliably extrapolated to the distinct environmental and evolutionary contexts of the tropical and arid hosts that dominate the global flora. Current estimates of global fungal richness may therefore require significant upward revision as we expand sampling beyond the few dominant plant hosts currently represented in the literature.\u003c/p\u003e\n\u003cp\u003eFrom \u0026ldquo;Study-Level Negatives\u0026rdquo; to \u0026ldquo;Taxon-Level Ubiquity\u0026rdquo;\u003c/p\u003e\n\u003cp\u003eA key challenge in testing ubiquity is the interpretation of \u0026ldquo;absence\u0026rdquo; reports. Our screening pipeline flagged 102 abstracts that indicated absence. Manual validation clarified that these were either irrelevant (40), simple false positives (i.e., abstracts incorrectly classified as \u0026lsquo;Absence\u0026rsquo; by the model; 46), or reports of mixed results (15) where endophytes were found in some tissues but not others or in some plant individuals but not others.\u003c/p\u003e\n\u003cp\u003eWe identified only a single genuine \u0026ldquo;study-level absence\u0026rdquo; in the literature: a study reporting no fungal endophytes in \u003cem\u003ePhragmites australis\u003c/em\u003e\u003csup\u003e16\u003c/sup\u003e. However, this absence is contradicted by numerous other studies that confirm diverse fungal endophyte communities in the same host species(e.g., \u003csup\u003e17\u0026ndash;19\u003c/sup\u003e). This specific case perfectly illustrates our core distinction: while individual studies may fail to detect endophytes (\u0026ldquo;study-level negatives\u0026rdquo;), \u0026ldquo;taxon-level ubiquity\u0026rdquo; holds true. We found no evidence of a plant taxon that is consistently free of fungal endophytes.\u003c/p\u003e\n\u003cp\u003eLimitations of Scope\u003c/p\u003e\n\u003cp\u003eOur analysis relies on metadata and claims extracted from scientific abstracts. While this approach enables a scale of analysis impossible with manual full-text review, it inherently produces a conservative dataset, as abstracts may not explicitly list every host species or collection site. However, given that coverage stands at only 0.8% of plant species, even a substantial increase in species recovery from full texts would not alter the fundamental conclusion: the vast majority of global plant biodiversity remains completely uninvestigated.\u003c/p\u003e\n\u003cp\u003eA critical factor influencing the published record is publication bias, often termed the \u0026apos;file drawer problem\u0026apos;\u003csup\u003e20\u003c/sup\u003e. In the context of the ubiquity paradigm, this bias is likely circular: failure to detect endophytes is often attributed to methodological error rather than biological absence, discouraging the publication of negative results. Thus, our \u0026lsquo;zero verifiable absences\u0026rsquo; finding reflects the consensus of the published record, which is itself shaped by the assumption of ubiquity.\u003c/p\u003e\n\u003cp\u003eFinally, our restriction to English-language abstracts imposes a linguistic filter on the map of knowledge. Consequently, the \u0026lsquo;zero\u0026rsquo; counts observed in many nations in the Global South (e.g., across substantial portions of Africa) likely reflect a compound failure: a genuine scarcity of funded biodiversity monitoring, the exclusion of non-English records within indexed databases, and a systemic indexing bias that omits regional journals entirely.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis concentration of research has profound practical consequences. By prioritizing the Global North, we are effectively ignoring the tropical and arid biodiversity hotspots of the Global South\u003csup\u003e21\u003c/sup\u003e, regions where plants face some of the most extreme environmental pressures\u003csup\u003e22\u003c/sup\u003e. This geographic bias is particularly critical given the contrasting biogeography of fungal guilds. While the diversity of key soil guilds, particularly ectomycorrhizal fungi (definitionally excluded from being endophytes), peaks in temperate and boreal zones\u003csup\u003e23\u003c/sup\u003e, foliar endophytes appear to follow the canonical latitudinal diversity gradient, reaching peak richness in the tropics\u003csup\u003e7\u003c/sup\u003e. By concentrating 77% of research in the Global North, the field has inadvertently aligned its sampling effort with the diversity centers of mycorrhizae, while systematically ignoring the hyperdiverse centers of endophytic fungi.\u003c/p\u003e\n\u003cp\u003eThis reframes the research bias from a simple academic gap to a significant bottleneck in global bioprospecting\u003csup\u003e5,24\u003c/sup\u003e. The cost of this omission is illustrated by recent successes in under-sampled, extreme environments; for instance, endophytes from the Antarctic angiosperm \u003cem\u003eColobanthus quitensis\u003c/em\u003e have yielded unique metabolic pathways with potential applications in neurodegenerative disease treatment\u003csup\u003e25\u003c/sup\u003e. Such discoveries confirm that the 99% of plant hosts currently excluded from our limited, sampling biased toward the Global North represent a vast, untapped reservoir of functional biology.\u003c/p\u003e\n\u003cp\u003eBeyond quantifying this bias, our analysis provides a corrective tool: a data-driven roadmap of the plant taxa and geographic regions that define the \u0026lsquo;missing\u0026rsquo; majority of the literature. Provided in the Supplementary Information, this resource identifies priority targets, such as the phyla Charophyta and Bryophyta and nations like Tonga, where directed sampling is most likely to maximize novelty. Having established the global ubiquity of fungal endophytes, the field must now pivot from verifying their presence to exploring their function. By redirecting research toward these neglected lineages and regions, we can unlock the next generation of climate-resilient symbioses and biotechnologies hidden within the world\u0026rsquo;s most threatened ecosystems.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003eOur methodology followed a three-phase computational workflow, beginning with the full dataset of 21,891 abstracts.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePhase 1: High-Sensitivity Screening. We developed a two-stage machine learning pipeline to first isolate relevant primary research (n=19,447) from the full dataset (n=21,891) and then classify these abstracts for potential absence claims.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePhase 2: Human Validation. We manually reviewed all abstracts flagged as potential absences (n=102) to verify the \u0026lsquo;zero absence\u0026rsquo; finding.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ePhase 3: Extraction \u0026amp; Mapping. We applied a validated extraction pipeline to the relevant corpus (n=19,447). To ensure our bias analysis focused strictly on endophytes, we excluded abstracts focused solely on mycorrhizal fungi, resulting in a final analytical dataset of 19,199 abstracts for taxonomic and geographic mapping.\u003c/p\u003e\n\u003ch3\u003e1. Machine Learning Classification as a High-Sensitivity Screen\u003c/h3\u003e\n\u003cp\u003eFor phase 1 of our workflow, we developed a two-stage machine learning pipeline. The first stage screened for relevance, resulting in the 19,447 abstract dataset. The second stage classified relevant abstracts as either indicating presence or absence of fungal endophytes in the studied plants. For this classification, we faced an extreme imbalanced class problem: our manually curated 876-abstract training dataset contained 346 \u0026lsquo;Presence\u0026rsquo; examples but only a single \u0026lsquo;Absence\u0026rsquo; example (0.11%). To address this scarcity, we employed data augmentation techniques, generating synthetic abstracts representing plausible absence claims using large language models (see Supplementary Methods). These synthetic data were used solely to define the linguistic boundaries during training and were not part of the analyzed dataset.\u003c/p\u003e\n\u003cp\u003eTo design a high-sensitivity screening tool, we used a weighted ensemble of two machine learning models: a Support Vector Machine (a model that finds the optimal boundary separating two groups\u003csup\u003e26\u0026ndash;28\u003c/sup\u003e) and a regularized logistic regression (a model that automatically selects the most important predictive words\u003csup\u003e29\u0026ndash;31\u003c/sup\u003e). As documented in our model testing, we intentionally weighted the model to prioritize absence detection, creating a screen that would flag all potential candidates for manual review.\u003c/p\u003e\n\u003ch3\u003e2. Pipeline Validation: Absence and Manual Review\u003c/h3\u003e\n\u003cp\u003eWe implemented a rigorous, multi-step process to validate all potential \u0026ldquo;Absence\u0026rdquo; classifications. First, our Machine Learning (ML) screen flagged 89 candidates. Second, an independent rule-based string detection algorithm flagged 12 additional unique candidates. Combined with the inclusion of our single positive control from the training set, this created a final set of 102 abstracts for manual validation.\u003c/p\u003e\n\u003cp\u003eWe manually reviewed all of these 102 abstracts. The results, detailed in the Supplementary Methods, confirmed that zero represented a true, taxon-level absence.\u003c/p\u003e\n\u003ch3\u003e3. Data Extraction Pipeline and Validation\u003c/h3\u003e\n\u003cp\u003eWe built a modular data extraction pipeline to parse taxonomic, geographic, and methodological information. The robustness of every component, from species name extraction to geographic coordinate parsing, was verified using a comprehensive, automated test suite. This suite evaluated core functionality, accuracy, synonym resolution, and error handling, ensuring high fidelity in the extracted data. After taxonomic information was attached to taxon names in the abstracts, any taxa listed as mycorrhizal in FungalTraits\u003csup\u003e32\u003c/sup\u003e and any taxa with the phylum Glomeromycota were removed. Although mycorrhizal fungi inhabit plant tissues, they represent a distinct, well-characterized functional guild separate from the general \u0026ldquo;endophyte\u0026rdquo; paradigm tested here\u003csup\u003e2\u003c/sup\u003e; therefore, abstracts focusing exclusively on mycorrhizal fungi were excluded to avoid confounding the results on endophytic fungi.\u003c/p\u003e\n\u003ch3\u003e4. Data Aggregation\u003c/h3\u003e\n\u003cp\u003eTaxonomic names were standardized, and synonyms were resolved against the Global Biodiversity Information Facility (GBIF) backbone\u003csup\u003e33\u003c/sup\u003e to ensure accurate aggregation of species counts. We used the Paleobiology Database\u003csup\u003e34\u003c/sup\u003e to exclude extinct taxa from the GBIF backbone. Geographic information was aggregated at the country and region level. For regional analysis, countries were classified as \u0026lsquo;Global North\u0026rsquo; or \u0026lsquo;Global South\u0026rsquo; based on the World Bank\u0026rsquo;s income group classifications (e.g., \u0026lsquo;High income\u0026rsquo; vs. \u0026lsquo;Low/Middle income\u0026rsquo;)\u003csup\u003e14\u003c/sup\u003e.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eAuthor Contributions\u003c/p\u003e\n\u003cp\u003eB.B. conceived the study, designed the methodology, performed the data analysis and machine learning, and wrote the manuscript. N.P.M., N.C.J. and C.A.G. contributed to the final manuscript text. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003eCompeting Interests\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003eFunding Statement\u003c/p\u003e\n\u003cp\u003eThis work was supported by National Science Foundation (NSF) MacroSystems grant DEB-1340852 (C.A.G.), US Department of Energy program in Systems Biology Research to Advance Sustainable Bioenergy Crop Development (DE-FOA-0002214)(N.C.J.), the Lucking Family Professorship (C.A.G.), the ARCS Foundation (B.M.B.), the Presidential Fellowship Program at NAU (B.M.B.), the Arizona Mushroom Society (B.M.B.), the Support for Graduate Students program at NAU (B.M.B.), and the American Association of University Women\u0026rsquo;s American Dissertation Fellowship (B.M.B.).\u003c/p\u003e\n\u003cp\u003eData and Code Availability\u003c/p\u003e\n\u003cp\u003eThe full code for the data analysis pipeline, including all R scripts used for machine learning, validation, and figure generation, is available at DOI: 10.5281/zenodo.17770500\u003csup\u003e35\u003c/sup\u003e. The final abstract dataset and model outputs generated during this study are available at DOI: 10.5281/zenodo.17770522\u003csup\u003e36\u003c/sup\u003e.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eKusari, S., Hertweck, C. \u0026amp; Spiteller, M. Chemical Ecology of Endophytic Fungi: Origins of Secondary Metabolites. \u003cem\u003eChem. Biol.\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 792\u0026ndash;798 (2012).\u003c/li\u003e\n\u003cli\u003eCosner, J. \u003cem\u003eet al.\u003c/em\u003e Fungal endophytes. \u003cem\u003eCurr. Biol.\u003c/em\u003e \u003cstrong\u003e35\u003c/strong\u003e, R904\u0026ndash;R910 (2025).\u003c/li\u003e\n\u003cli\u003ePetrini, O. Fungal Endophytes of Tree Leaves. in \u003cem\u003eMicrobial Ecology of Leaves\u003c/em\u003e (eds Andrews, J. H. \u0026amp; Hirano, S. S.) 179\u0026ndash;197 (Springer New York, New York, NY, 1991). doi:10.1007/978-1-4612-3168-4_9.\u003c/li\u003e\n\u003cli\u003eRodriguez, R. J., White Jr, J. F., Arnold, A. E. \u0026amp; Redman, R. S. Fungal endophytes: diversity and functional roles. \u003cem\u003eNew Phytol.\u003c/em\u003e \u003cstrong\u003e182\u003c/strong\u003e, 314\u0026ndash;330 (2009).\u003c/li\u003e\n\u003cli\u003eStrobel, G. A. Endophytes as sources of bioactive products. \u003cem\u003eMicrobes Infect.\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 535\u0026ndash;544 (2003).\u003c/li\u003e\n\u003cli\u003eStone, J. K., Bacon, C. W. \u0026amp; White, F. An Overview of Endophytic Microbes: Endophytism Defined. in \u003cem\u003eMicrobial Endophytes\u003c/em\u003e 1\u0026ndash;28 (CRC Press, 2000).\u003c/li\u003e\n\u003cli\u003eArnold, A. \u0026amp; Lutzoni, F. Diversity and host range of foliar fungal endophytes: Are tropical leaves biodiversity hotspots? \u003cem\u003eEcology\u003c/em\u003e \u003cstrong\u003e88\u003c/strong\u003e, 541\u0026ndash;549 (2007).\u003c/li\u003e\n\u003cli\u003eSchulz, B. \u0026amp; Boyle, C. The endophytic continuum. \u003cem\u003eMycol. Res.\u003c/em\u003e \u003cstrong\u003e109\u003c/strong\u003e, 661\u0026ndash;686 (2005).\u003c/li\u003e\n\u003cli\u003eU\u0026rsquo;Ren, J. M., Lutzoni, F., Miadlikowska, J., Laetsch, A. D. \u0026amp; Arnold, A. E. Host and geographic structure of endophytic and endolichenic fungi at a continental scale. \u003cem\u003eAm. J. Bot.\u003c/em\u003e \u003cstrong\u003e99\u003c/strong\u003e, 898\u0026ndash;914 (2012).\u003c/li\u003e\n\u003cli\u003eSagi, O. \u0026amp; Rokach, L. Ensemble learning: A survey. \u003cem\u003eWIREs Data Min. Knowl. Discov.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, e1249 (2018).\u003c/li\u003e\n\u003cli\u003eShah, M. \u003cem\u003eet al.\u003c/em\u003e Theoretical Evaluation of Ensemble Machine Learning Techniques. in \u003cem\u003e2023 5th International Conference on Smart Systems and Inventive Technology (ICSSIT)\u003c/em\u003e 829\u0026ndash;837 (IEEE, Tirunelveli, India, 2023). doi:10.1109/ICSSIT55814.2023.10061139.\u003c/li\u003e\n\u003cli\u003eMartin, F. M., Uroz, S. \u0026amp; Barker, D. G. Ancestral alliances: Plant mutualistic symbioses with fungi and bacteria. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e356\u003c/strong\u003e, eaad4501 (2017).\u003c/li\u003e\n\u003cli\u003eStrullu‐Derrien, C., Selosse, M., Kenrick, P. \u0026amp; Martin, F. M. The origin and evolution of mycorrhizal symbioses: from palaeomycology to phylogenomics. \u003cem\u003eNew Phytol.\u003c/em\u003e \u003cstrong\u003e220\u003c/strong\u003e, 1012\u0026ndash;1030 (2018).\u003c/li\u003e\n\u003cli\u003eWorld Bank. World Bank Country and Lending Groups. (2025).\u003c/li\u003e\n\u003cli\u003eHawksworth, D. L. \u0026amp; L\u0026uuml;cking, R. Fungal Diversity Revisited: 2.2 to 3.8 Million Species. \u003cem\u003eMicrobiol. Spectr.\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 5.4.10 (2017).\u003c/li\u003e\n\u003cli\u003eLambert, A. \u0026amp; Casagrande, R. No evidence of fungal endophytes in native and exotic Phragmites australis. \u003cem\u003eNortheast. Nat.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 561\u0026ndash;568 (2006).\u003c/li\u003e\n\u003cli\u003eWirsel, S. G. R., Leibinger, W., Ernst, M. \u0026amp; Mendgen, K. Genetic diversity of fungi closely associated with common reed. \u003cem\u003eNew Phytol.\u003c/em\u003e \u003cstrong\u003e149\u003c/strong\u003e, 589\u0026ndash;598 (2001).\u003c/li\u003e\n\u003cli\u003eAngelini, P. \u003cem\u003eet al.\u003c/em\u003e The endophytic fungal communities associated with the leaves and roots of the common reed (Phragmites australis) in Lake Trasimeno (Perugia, Italy) in declining and healthy stands. \u003cem\u003eFungal Ecol.\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 683\u0026ndash;693 (2012).\u003c/li\u003e\n\u003cli\u003eClay, K., Shearin, Z., Bourke, K., Bickford, W. \u0026amp; Kowalski, K. Diversity of fungal endophytes in non-native Phragmites australis in the Great Lakes. \u003cem\u003eBiol. Invasions\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 2703\u0026ndash;2716 (2016).\u003c/li\u003e\n\u003cli\u003eRosenthal, R. The file drawer problem and tolerance for null results. \u003cem\u003ePsychol. Bull.\u003c/em\u003e \u003cstrong\u003e86\u003c/strong\u003e, 638\u0026ndash;641 (1979).\u003c/li\u003e\n\u003cli\u003eMyers, N., Mittermeier, R. A., Mittermeier, C. G., Da Fonseca, G. A. B. \u0026amp; Kent, J. Biodiversity hotspots for conservation priorities. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e403\u003c/strong\u003e, 853\u0026ndash;858 (2000).\u003c/li\u003e\n\u003cli\u003eSeneviratne, S. I. \u0026amp; Zhang, X. Weather and Climate Extreme Events in a Changing Climate. in \u003cem\u003eClimate Change 2021 \u0026ndash; The Physical Science Basis: Working Group I Contribution to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change\u003c/em\u003e (Cambridge University Press, 2023). doi:10.1017/9781009157896.\u003c/li\u003e\n\u003cli\u003eTedersoo, L. \u003cem\u003eet al.\u003c/em\u003e Global diversity and geography of soil fungi. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e346\u003c/strong\u003e, 1256688 (2014).\u003c/li\u003e\n\u003cli\u003eNewman, D. J. \u0026amp; Cragg, G. M. Natural Products As Sources of New Drugs over the 30 Years from 1981 to 2010. \u003cem\u003eJ. Nat. Prod.\u003c/em\u003e \u003cstrong\u003e75\u003c/strong\u003e, 311\u0026ndash;335 (2012).\u003c/li\u003e\n\u003cli\u003eBertini, L. \u003cem\u003eet al.\u003c/em\u003e Biodiversity and Bioprospecting of Fungal Endophytes from the Antarctic Plant Colobanthus quitensis. \u003cem\u003eJ. Fungi\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, (2022).\u003c/li\u003e\n\u003cli\u003eJoachims, T. A statistical learning learning model of text classification for support vector machines. in \u003cem\u003eProceedings of the 24th annual international ACM SIGIR conference on Research and development in information retrieval\u003c/em\u003e 128\u0026ndash;136 (ACM, New Orleans Louisiana USA, 2001). doi:10.1145/383952.383974.\u003c/li\u003e\n\u003cli\u003eDrake, J. M., Randin, C. \u0026amp; Guisan, A. Modelling ecological niches with support vector machines. \u003cem\u003eJ. Appl. Ecol.\u003c/em\u003e \u003cstrong\u003e43\u003c/strong\u003e, 424\u0026ndash;432 (2006).\u003c/li\u003e\n\u003cli\u003eMammone, A., Turchi, M. \u0026amp; Cristianini, N. Support vector machines. \u003cem\u003eWIREs Comput. Stat.\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, 283\u0026ndash;289 (2009).\u003c/li\u003e\n\u003cli\u003eGenkin, A., Lewis, D. D. \u0026amp; Madigan, D. Large-Scale Bayesian Logistic Regression for Text Categorization. \u003cem\u003eTechnometrics\u003c/em\u003e \u003cstrong\u003e49\u003c/strong\u003e, 291\u0026ndash;304 (2007).\u003c/li\u003e\n\u003cli\u003eSalehi, F., Abbasi, E. \u0026amp; Hassibi, B. The Impact of Regularization on High-dimensional Logistic Regression. Preprint at https://doi.org/10.48550/arXiv.1906.03761 (2019).\u003c/li\u003e\n\u003cli\u003eIchsan, A., Riyadi, S. \u0026amp; Pardede, D. Analysis of Logistic Regression Regularization in Wild Elephant Classification with VGG-16 Feature Extraction. \u003cem\u003eJ. Comput. Netw. Archit. High Perform. Comput.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 783\u0026ndash;793 (2024).\u003c/li\u003e\n\u003cli\u003eP\u0026otilde;lme, S. \u003cem\u003eet al.\u003c/em\u003e FungalTraits: a user-friendly traits database of fungi and fungus-like stramenopiles. \u003cem\u003eFungal Divers.\u003c/em\u003e \u003cstrong\u003e105\u003c/strong\u003e, 1\u0026ndash;16 (2020).\u003c/li\u003e\n\u003cli\u003eRegistry-Migration.Gbif.Org. GBIF Backbone Taxonomy. GBIF Secretariat https://doi.org/10.15468/39OMEI (2023).\u003c/li\u003e\n\u003cli\u003eUhen, M. D. \u003cem\u003eet al.\u003c/em\u003e Paleobiology Database User Guide Version 1.0. \u003cem\u003ePaleoBios\u003c/em\u003e \u003cstrong\u003e40\u003c/strong\u003e, (2023).\u003c/li\u003e\n\u003cli\u003eBock, B. M. beabock/Endo-Review: endo-review: a ML pipeline to review fungal endophyte ubiquity. Zenodo https://doi.org/10.5281/ZENODO.17770500 (2025).\u003c/li\u003e\n\u003cli\u003eBock, B. M. Endo-Review: Data, Models, and Results. Zenodo https://doi.org/10.5281/ZENODO.17770522 (2025).\u003c/li\u003e\n\u003cli\u003eChawla, N. V., Bowyer, K. W., Hall, L. O. \u0026amp; Kegelmeyer, W. P. SMOTE: Synthetic Minority Over-sampling Technique. \u003cem\u003eJ. Artif. Intell. Res.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 321\u0026ndash;357 (2002).\u003c/li\u003e\n\u003cli\u003eKuhn, M. caret: Classification and Regression Training. 7.0\u0026ndash;1 https://doi.org/10.32614/CRAN.package.caret (2007).\u003c/li\u003e\n\u003cli\u003eRaasveldt, M. \u0026amp; M\u0026uuml;hleisen, H. DuckDB: an Embeddable Analytical Database. in \u003cem\u003eProceedings of the 2019 International Conference on Management of Data\u003c/em\u003e 1981\u0026ndash;1984 (ACM, Amsterdam Netherlands, 2019). doi:10.1145/3299869.3320212.\u003c/li\u003e\n\u003cli\u003eR Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing (2023).\u003c/li\u003e\n\u003cli\u003eWickham, H. \u003cem\u003eGgplot2: Elegant Graphics for Data Analysis\u003c/em\u003e. (Springer international publishing, Cham, 2016).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8244775/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8244775/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"The paradigm that ‘all plants harbor fungal endophytes’ is a widely accepted axiom in plant ecology, yet this claim has never been tested. Here, we use a machine learning pipeline to screen 21,891 research abstracts spanning nearly a century. We developed a high-sensitivity screening tool to find and manually validate all potential claims of endophyte absence. This comprehensive validation revealed no evidence of a verifiable endophyte-free plant taxon within the entire scientific literature surveyed. This finding confirms that endophyte ubiquity is a robust pattern wherever plants have been studied. However, we show that this “ubiquitous” paradigm is a generalization from an extraordinarily biased sample: analysis of the 19,119 relevant primary studies reveals that only 0.8% of described plant species have been examined for fungal endophytes, with 77% of this research concentrated in the Global North. Our analysis reframes a foundational paradigm, revealing that our understanding of a “global” symbiosis is based on a tiny fraction of biodiversity. This sampling bias represents a critical bottleneck in biodiversity research, obscuring the discovery of novel biotechnologies and climate-resilient symbioses in the world’s most threatened ecosystems.","manuscriptTitle":"The Sparsely Sampled Ubiquity of Global Fungal Endophytes","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-02 13:17:59","doi":"10.21203/rs.3.rs-8244775/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"031d621d-5c54-4239-b011-b063106babc9","owner":[],"postedDate":"December 2nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":58847410,"name":"Biological sciences/Ecology/Macroecology"},{"id":58847411,"name":"Biological sciences/Ecology/Conservation biology"},{"id":58847412,"name":"Biological sciences/Ecology/Biodiversity"},{"id":58847413,"name":"Biological sciences/Ecology/Biogeography"},{"id":58847414,"name":"Biological sciences/Ecology/Microbial ecology"}],"tags":[],"updatedAt":"2025-12-03T17:01:49+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-02 13:17:59","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8244775","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8244775","identity":"rs-8244775","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}