henhao
Sign in
EcologybioRxiv
Preprint

Fast pace of life is associated with host status in small mammals, despite pervasive surveillance bias

David SimonsiD, Ricardo RiveroiD, Grant RickardiD, Ana Martinez-Checa GuioteiD, Harry Gordon, David W. ReddingiD, Stephanie N. SeifertiD

Abstract

Small mammals are the principal hosts of arenaviruses and hantaviruses, yet the determinants of host status remain contested because global surveillance is non-random. We harmonised 729 studies comprising 695,000 diagnostic assays across 637 species and asked whether host status is predictable once surveillance effort is accounted for. Surveillance intensity was predicted by night-time light intensity and accessibility, but not local host richness, and sampling effort was phylogenetically clustered (Pagel's {lambda} = 0.43). Forty-six percent of rodent genera have never been sampled. In Bayesian phylogenetic mixed models of 43,677 host-virus pairs, faster-lived species tended to be detected as hosts more often, though the credible interval included zero (posterior mean = -0.125, 95% CrI [-0.29, 0.04], pd = 93.2%); the estimate was comparable when restricted to active infection. Synanthropy was independently associated with host status, obligate commensals showing approximately twice the odds. Predictions discriminated hosts from non-hosts in withheld continents (AUC 0.81-0.88), scored at observed sampling effort. Host and virus phylogenies were congruent in both families, though most associations departed from strict co-divergence, indicating host-switching alongside co-evolutionary history. Host phylogenetic diversity accumulated without saturating, and arenavirus hosts were discovered in a more phylogenetically clustered order than expected by chance, consistent with surveillance concentrating on relatives of known hosts. Projected globally, community host probability varied largely independently of species richness (R2 = 0.028), so diverse assemblages do not carry systematically elevated host probability. Surveillance bias is substantial and quantifiable, and host biology retains predictive value once it is accounted for.

PDFDOICodeData
Full-text digest ↓

02 /Specification & Classification

Click any tag to browse matching papers
Citations
1
OpenAlex Topics

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Sonnet 4.6 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (1 Oct 2026) · CC BY 4.0

Arenavirus and hantavirus surveillance in small mammals tracks night-lights and accessibility; after adjusting for effort, fast-lived and synanthropic species are somewhat more often detected as hosts.

Reservoir host lists for arenaviruses and hantaviruses come from uneven, convenience-driven sampling, which can distort or mask apparent links between host traits and host status. This work quantifies that bias and then tests life-history and synanthropy effects after adjusting for sampling effort and phylogeny.

Key results

  1. ShownIn the count model, sampling tracked night-light intensity and accessibility rather than richness; 46% of Rodentia and Eulipotyphla genera are unsampled and speciose tropical regions remain largely uncharacterised.

    In the ZINB district model, night-light intensity and accessibility were the strongest predictors of sampling intensity; richness and population density were not important in the count component. 46% of Rodentia and Eulipotyphla genera unsampled; Palearctic and Nearctic realms account for 71% of detected small mammals. · The Anthropogenic Filter: Quantifying…; Discussion

  2. SuggestedA fast pace of life (low PC1) is weakly and uncertainly associated with higher detected host status after adjusting for sampling effort and phylogeny; the active-infection-only model gives a similar estimate.

    Global model (N = 43,677 dyadic pairs): PC1 posterior mean = −0.125, 95% CrI [−0.29, 0.04], pd = 0.932. Active infection only: −0.17 [−0.40, 0.07], pd = 91.8%. Both credible intervals include zero. · Macroecological Associations with Host Status; Phylogenetic Dyadic Modelling

  3. SuggestedObligate synanthropy predicts host status independently of pace of life, but the interval only narrowly excludes zero and the subset is smaller.

    Synanthropy subset (N = 18,906 dyadic pairs): totally synanthropic vs non-synanthropic posterior mean = 0.762, 95% CrI [0.045, 1.46], about 2.2-fold increase in odds. PC1 estimate −0.147 in this subset vs −0.125 in the global model. · Macroecological Associations with Host Status; Discussion

  4. ShownModels trained on other continents discriminated hosts from non-hosts in each withheld continent, although observed sampling effort contributes to this performance.

    Leave-One-Region-Out AUC: 0.810 (Americas), 0.832 (Asia), 0.826 (Europe), 0.876 (Africa). Out-of-sample vs full-model predictions Spearman's ρ = 0.767. Genus-level overlap with training set: 8.8% Americas, 13.0% Africa, 26.7% Asia, 37.5% Europe. · Macroecological Associations with Host Status; Global Host Probability Mapping; Discussion

  5. SuggestedBoth families show significant lineage-level host-virus congruence, with link-level incongruence suggesting recurrent host switching; arenavirus host discovery was phylogenetically clustered and host breadth has not saturated.

    PACo: Arenaviridae (16 viruses, 41 hosts, 47 links) p < 0.001; Hantaviridae (30 viruses, 90 hosts, 141 links) p < 0.001. Arenavirus discovery curve area 13% below null median (10,000 permutations), one-tailed p = 0.037; hantavirus 4% below, p = 0.12. · Evolutionary Congruence and Host-Switching; Host Discovery Is Phylogenetically Clustered and…; Analysis of Evolutionary Congruence; Host Phylogenetic Diversity Accumulation and…

Methods, data and reproducibility

A systematic review of surveillance records fed a ZINB model of sampling effort and Bayesian phylogenetic dyadic GLMMs (brms/Stan) of host status on pace-of-life PC1 and synanthropy, validated by leave-one-region-out cross-validation, plus PACo cophylogeny and host phylogenetic diversity accumulation analyses.

  • Systematic review (PubMed, Web of Science; January 1960 to August 2023) of arenavirus and hantavirus surveillance in Rodentia and Eulipotyphla; ArHa database · Small mammals (rodents, shrews) and viruses · n = approx. 561,000 individuals screened; 637 host species · 10.5281/zenodo.20523966
  • Global dyadic model dataset (host-virus pairs, PC1 pace of life from imputed COMBINE traits) · Rodents and shrews · n = 43,677 dyadic pairs; 69 pathogens; 633 species
  • Synanthropy subset with detailed synanthropy data · Rodents and shrews · n = 18,906 dyadic pairs; 274 species
  • Complete-case subset with empirical, non-imputed life-history traits · Rodents and shrews · n = 14,697 dyadic pairs; 213 species
  • Life-history projection to all small mammal species with trait data, mapped on GADM Level 2 districts at 20km resolution · Small mammals · n = 2,766 species; 37,676 of 38,367 districts retained
  • Host-virus PCR-positive association matrices for cophylogeny, Arenaviridae and Hantaviridae · Small mammals and viruses · n = Arena: 16 viruses, 41 hosts, 47 links; Hanta: 30 viruses, 90 hosts, 141 links
Methods & tools
brms (2.22), Stan, mgcv (1.9-3), Rphylopars (0.3.10), taxize (0.10), COMBINE database, PACo, ParaFit, BEAST X (HIPSTR trees), terra (1.8-60), biscale (1.1), exactextractr (0.10)
Code
https://github.com/DidDrog11/arha-macroecology
Data & accessions
https://doi.org/10.5281/zenodo.20523966
Software & environment
R 4.2.3; taxize 0.10; brms 2.22; mgcv 1.9-3; Rphylopars 0.3.10; terra 1.8-60; biscale 1.1; exactextractr 0.10; MDD v1.11; IUCN Red List v2024-2; ICTV 2024 release.
Reproducibility gaps
Repository does not host environmental spatial layers or IUCN range polygons; R scripts fetch and harmonise them from original public sources.
Limitations (5)
  • Seroconversion may reflect transient or dead-end infection, and over 67% of historical host-virus associations rest only on serological or fragment-based evidence, which may inflate apparent host ranges. · stated by the authors
  • Trait coverage is sparse in the tropics (imputation 58.5% Afrotropic, 69.5% Neotropic vs 6.1% Palearctic-North), so host probability there may be underestimated; 7 species with confirmed positives were excluded. · stated by the authors
  • Total individuals tested is an imperfect effort measure because testing is targeted by viral family (e.g., 87.2% Hantaviridae in the United States, 100% Arenaviridae in Nigeria and Sierra Leone). · stated by the authors
  • Whether the fast pace-of-life signal reflects viral tolerance cannot be determined from surveillance data; longitudinal or experimental infection studies would be required. · stated by the authors
  • No effort-only baseline model is reported alongside the LORO AUCs, so the added predictive value of host biology beyond sampling effort is not quantified. · noted in this summary

Who should read:Read this if you model reservoir hosts from biased surveillance data and want worked Bayesian phylogenetic approaches for adjusting sampling effort, with cross-region validation.

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISGibb '21Cobos '25Albery '20Dubrulle '23Walsh '18Song '25
66% semantic matchEcology
R. Gibb, G. F. Albery, N. F. Mollentze+5 more·bioRxiv·Aug 10, 2021