What models, genomics, and trial frameworks suggest about containment, spread, and real-world feasibility
This version (Version 2) was created on 16 April 2026.
Engineered gene drives (EGD) promise targeted, one-time interventions that could suppress invasive rodents on islands or reduce malaria transmission by mosquitoes. The first question is always the same: where do you start without risking uncontrolled spread? Islands keep surfacing as the answer. They are small, well-mapped and surrounded by water. Water acts as a barrier. Globally, islands represent less than 6% of land area but host a disproportionate share of biodiversity, including many endemic species highly vulnerable to invasive species impacts. Early trials there could show results before the technology reaches the continents. At least theoretically.
In conservation, the focus falls on invasive rodents. Hay and Guo (2022) show that mice and rats drive most recent extinctions on islands, especially of ground-nesting birds and other small native animals, by eating eggs and chicks and destroying nests. Invasive species are implicated in around 60% of global vertebrate extinctions, with rodents among the most damaging on islands (IPBES, 2019). Rodenticide campaigns work on small islands but scale poorly: costs rise, logistics become difficult and non-target species suffer. Large-scale eradication campaigns can cost millions of dollars and often require repeated applications, particularly on inhabited or topographically complex islands. EGD suppression appears as the next logical step in the sequence of control methods. It targets the pest directly, requires no repeated applications, and offers lasting suppression after a single release.
Brodie and co-authors (2025) place the same idea in policy terms. They describe “synthetically assisted conservation” as an extension of existing tools: conservation translocation for species recovery and integrated pest management for invasives. Engineered gene drives fit alongside sterile insect technique and classical biocontrol. This shift matters. The question changes from “should we use gene drives?” to “under what conditions do they belong in current management frameworks?” Islands emerge not only as test sites but as priority locations where native species face acute pressure and where targeted genetic tools can complement methods with known limitations.
The same logic appears in mosquito control. Vorsino and colleagues (2025) discuss Aedes mosquitoes and arbovirus risk in Pacific Island Countries and Territories. Low-lying atolls look attractive for genetic biocontrol: remote, small land area. Yet boat and air links connect them to larger economies. Surveillance and control capacity vary. The result is not a closed system but a connected network of islands. Isolation on a map thus does not necessarily mean isolation in practice.
Across the scientific literature, islands serve as the bridge from concept to field application. Authors move beyond “could gene drives work?” to “where and how should early trials happen?” Recent modelling studies suggest that gene drive systems could achieve substantial population suppression under favourable conditions, with predicted reductions of up to 70–90% or more in target populations depending on design and context. This article follows that progression. I examine what the research says regarding ideal trial sites, what models and genomic data predict about spread and localisation in islands, the management realities of trials on islands, and the governance needed for responsible deployment.
Selecting islands for gene drive field trials: criteria for isolation, containment and feasibility
How might researchers select an island for the first open-field trial of an EGD mosquito aimed at malaria control? What parameters would define a site capable of delivering clear, interpretable data with minimal risk of unintended spread? Lanzaro and colleagues (2021) tackle these questions head-on in a landmark framework focused on population modification strategies using threshold-independent drives to spread parasite-blocking effectors through Anopheles vectors.
The authors stress that early gene drive Phase 2 field trials, as outlined in endorsed WHO pathways, shift the emphasis from laboratory or caged evaluations to real-world conditions. Phase 2 trials measure key entomological endpoints: survival of the engineered mosquitoes, their mating competitiveness against wild types, the spread of the gene drive construct, and its genetic stability over time. These studies aim to generate data that cannot be fully captured in confined settings, yet they demand strong containment to ensure measurements remain accurate and risks stay low.
To meet these demands, the authors propose physical islands as prime candidates, drawing on island biogeography theory to prioritise geographic and genetic isolation, moderate size, topographic complexity, low anopheline species richness and limited human-mediated dispersal. They present a structured set of criteria evaluated before social or legal factors (Figure 1).
Geographic isolation limits gene flow in both directions, reducing immigration of wild mosquitoes that could dilute gene drive effects or emigration that raises regulatory concerns. Genetic isolation is confirmed through population genomics, with oceanic islands showing clear divergence from mainland populations (Figure 2).
The evaluation covered 22 African islands and groups (Figure 3). Many fell short: continental and lakeside sites often lacked sufficient isolation or exhibited high complexity, while some oceanic ones were disqualified for excessive size (Madagascar), insufficient area (Annobón), or absence of target vectors. São Tomé & Príncipe and the Comoros archipelago (Grande Comore, Anjouan, Mohéli, Mayotte) emerged as leading candidates, favoured for robust isolation, suitable scale to exceed typical mosquito flight ranges (assumed up to 10 km daily), environmental heterogeneity via topography, fewer non-target Anopheles species (simplifying identification and reducing hybridisation risks), and low air/sea traffic.
Jones and colleagues (2021) broaden the discussion by considering islands in the context of cluster-randomised trials (CRTs), the gold-standard design for area-wide vector interventions where effects are assessed at the community or cluster level rather than the individual level. In CRTs, clusters (such as villages or islands) are randomly assigned to intervention or control arms to minimise bias from confounding variables. Islands offer advantages here: moderate human movement and short vector dispersal distances. Dispersal is typically limited to 0.5–1 km, with occasional movement beyond 7 km, while some modelling assumptions extend daily flight ranges up to 10 km, shaping how island size influences containment potential.
For self-propagating EGD strains, such isolation could support containment if unexpected issues arise. Still, the authors note limitations: archipelagos may restrict the number of available independent clusters needed for statistical power, small island populations can constrain sample sizes, and baseline heterogeneity (for example, in disease prevalence) may require more clusters to detect effects reliably. Outcomes from island-based trials therefore warrant cautious interpretation when generalising to other contexts, consistent with calls for multi-site evidence across varied entomological settings.
This conceptual blueprint, based on careful geographic and biological concepts, invites the next layer of inquiry: what do mathematical models and genomic analyses reveal about how gene drives might actually spread, localise, or spill over in these island–mainland configurations?
Gene drive spread on islands: insights from modelling and population genomics
For readers seeking a broader foundation in how mathematical and computational models forecast gene drive behaviour, I recommend first consulting From lab to landscape: How models predict gene drive spread. With that groundwork in place, the present section turns to the specific insights that emerge when models and genomic analyses are applied to spatially structured environments such as islands.
Rather than treating islands as sealed units, the research increasingly asks what happens when dispersal, migration, population structure and ecological realism are taken seriously. The answers differ markedly depending on the target taxon: models for invasive rodents tend to focus on complete eradication from small, geographically isolated islands, whereas those for malaria-vector mosquitoes explore partial suppression across larger, more heterogeneous and connected landscapes.
In the rodent literature, spatially explicit individual-based simulations have revealed both promise and caution. Birand et al. (2022) calibrated their stochastic model to a former invasive mouse population on Antipodes Island, modelling a homing gene drive against a fertility gene and a Y-linked X-shredding drive that biases sex ratios. Spatial models suggest that releasing as few as 128 engineered mice could suppress island populations of around 200,000 individuals, with eradication occurring over 4–8 years, although outcomes vary depending on spatial dynamics. Yet the pathway was rarely straightforward: some simulations showed a rapid, smooth collapse, while others produced prolonged chase dynamics (local wipe-out followed by comeback) from wild-type animals in nearby areas (Figure 4). These cycles highlight a recurring limitation in spatial models: when local drive spread outpaces recolonisation, eradication can stall, suggesting that panmictic (i.e., random mating within a breeding population) assumptions may overestimate success in fragmented habitats.
Gierus et al. (2022) pursued a different angle by embedding CRISPR elements into the naturally occurring mice “t” haplotype to create tCRISPR, a male meiotic drive that progressively depletes fertile females. Their spatially explicit simulations on a hypothetical island again pointed to eradication potential, supported by laboratory proof-of-concept for a split-drive format. The approach appears robust across parameter ranges, yet the authors openly note gaps: deme structure (i.e., the organisation of a species into small, local subpopulations that freely interbreed among themselves but have limited gene flow with other such groups), seasonal fluctuations, unanticipated fitness costs, and possible immune responses to Cas9 could all alter outcomes in ways that remain largely untested empirically.
Genomic data add an important layer by testing whether natural genetic differences could enforce localisation. Sudweeks et al. (2019) proposed the Locally Fixed Alleles (LFA) strategy, in which a homing gene drive targets a version of a gene that has become fixed on the island due to chance or founder effects but remains rare or absent elsewhere. Their deterministic models showed that an above-threshold release could suppress the island population while migration to a mainland kept drive frequencies below invasion thresholds, resulting in only transient, low-level exposure outside the target (Figure 5). Oh et al. (2021) then examined real invasive mouse populations on four small islands, using pooled whole-genome resequencing to identify thousands of fixed SNPs in genic regions suitable as targets. The reduced allelic diversity observed across sites supports the drift-driven fixation assumed by LFA, although the authors acknowledge trade-offs: targeting alleles that are very rare in source populations minimises spillover risk but narrows the pool of viable CRISPR sites.
Mosquito models, by contrast, deal with far larger and less isolated spatial scales. Pan and Champer (2022) examined continuous environments, where individuals mate locally but the species range extends widely, and found that drive fronts (they call it waves) propagate at speeds determined by conversion efficiency, fitness costs and dispersal. Some designs formed stable fronts; others stalled, revealing that continuous-space dynamics can produce qualitatively different behaviour from discrete or panmictic models. Kim et al. (2023) used simpler two-population setups to show how even modest unidirectional or bidirectional migration raises release thresholds and can prevent fixation or drive loss in the target population, framing islands as conceptually useful but vulnerable to leakage.
Hancock et al. (2024) took this further with high-resolution, geospatial simulations across sixteen 12,000 km² areas in West Africa. Their large-scale simulations predict 72–98% reductions in malaria vector populations when gene drives are deployed under favourable conditions. Spatial structure proved decisive: dense, interconnected subpopulations sustained suppression, whereas sparse or strongly seasonal landscapes permitted local recolonisation reminiscent of rodent chase dynamics. The work also demonstrated substantial additive benefits when drives were layered over existing interventions such as nets and vaccination, yet it left open questions about importation from outside modelled boundaries and the long-term stability of suppression in heterogeneous vector communities.
When confinement is considered directly, common themes appear. Dhole et al. (2018) compared self-limiting designs under migration, finding daisy-chain drives most efficient in isolation but most prone to spread between connected populations, while one-locus underdominance drives offered the tightest localisation. Greenbaum et al. (2021) showed that differential targeting (high drive frequency in the target, low in the non-target) is achievable with certain parameter combinations but fragile to rising migration rates, with rodents (lower conversion) more sensitive to dominance effects than mosquitoes. Willis and Burt (2021) proposed double-drive architectures that exploit pre-existing sequence differences, gaining robustness against resistance and permitting efficiency gains through sex-ratio distortion.
Across rodent and mosquito systems alike, the literature reveals that while models and genomic data can substantiate the feasibility of targeted EGD deployment and point to strategies for limiting potential spillover, they consistently flag the same core challenges: migration-driven leakage, spatial heterogeneity and the difficulty of achieving reliable localisation in anything but the most isolated contexts. The next section, therefore, turns to that next level of practical considerations of trial design, contamination control, vector behaviour and management frameworks as they appear in discussions of potential trials on islands.
Designing gene drive field trials on islands: logistics, monitoring and implementation
Models have done much of the heavy lifting so far: they have mapped out plausible pathways for spread, flagged spatial risks and suggested ways to contain or localise EGD systems on islands. I turn now to a different question: how might those predictions hold up when the first small-scale trials actually take place on real islands? The shift is from abstraction to implementation, where human presence, daily routines, boat traffic, seasonal weather and the need to layer new tools alongside established malaria control measures introduce complexities that simulations can only approximate.
Connolly and colleagues (2024) offer the most detailed roadmap for low-threshold gene drive trials in malaria-vector mosquitoes, grounded in expert discussions that build directly on the WHO guidance framework. They describe a phased testing pathway that starts in the laboratory, moves through the determination of genetic and entomological efficacy in controlled field settings, and eventually assesses epidemiological impacts and long-term surveillance (Figure 6). As low-threshold gene drives are designed to persist and spread, the authors propose adaptive trial designs: flexible protocols that allow incremental escalation of goals once early safety and genetic endpoints are met. This flexibility is presented as essential on islands, where containment cannot be absolute and outcomes may evolve over longer timescales than in short-term conventional trials.
Site selection remains a central preoccupation, especially for island contexts. Connolly et al. (2024) list five interlocking criteria:
- Operational viability: regulatory readiness, surveillance capacity, local expertise, accessibility, staff safety
- Species composition: risk of confounding from non-target Anopheles vectors or introgression into sibling species
- Statistical power: transmission intensity high enough to detect effects but not so overwhelming that impacts are masked
- Degree of isolation: geographic, seasonal, climatic, or genetic to slow outward gene flow and reduce inward confounding
- Potential for reversibility: theoretical mechanisms to halt or reverse drive activity
Islands feature prominently in these discussions because of their natural boundaries, yet the Connolly and colleagues (2024) is honest about the limits: regular air and sea links can transport mosquitoes or drive-carrying individuals to the mainland, high-altitude wind dispersal remains possible, and inward migration from nearby populations can dilute measured efficacy. Earlier proposals for oceanic islands such as São Tomé or the Comoros (echoed in Lanzaro et al. 2021 and others) are acknowledged, but so are the trade-offs: stronger isolation often comes with logistical hurdles, while continental lakeside islands (e.g., in Lake Victoria) might offer a compromise with greater ecological similarity to mainland sites.
Adaptive designs are suggested as one way to respond dynamically, perhaps starting with very small releases to monitor initial spread before scaling, yet the literature recognises that indefinite persistence complicates traditional placebo-controlled structures. Integration with existing interventions is another recurring point: gene drives are consistently positioned as complementary to insecticide-treated nets, indoor residual spraying and newer tools, potentially delivering additive reductions in transmission. On islands, where baseline malaria pressure can vary sharply with seasonality or tourism-related movement, careful layering becomes even more critical to avoid unintended interference or to maximise synergy.
Cross-references from EGD rodent-focused work reinforce similar concerns. Birand et al. (2022) and Gierus et al. (2022) show that even efficient drives can face delays from dispersal, polyandry or chase dynamics in spatially structured island populations, implying that trial monitoring must be intensive and long-term to capture such patterns. Pan and Champer (2022) and Kim et al. (2023) add that imperfect performance or modest migration can stall drive waves or elevate release thresholds, suggesting that island trials would benefit from pre-deployment field parameterisation of dispersal and structure. Oh et al. (2021) emphasise the value of genomic surveys to map local variation and inform construct design, practical groundwork that becomes indispensable when moving to real sites.
Models and criteria provide direction, but they cannot replicate the lived experience of island communities, fluctuating human behaviour or the day-to-day realities of surveillance and logistics. Decision-making is therefore framed as inherently iterative and collective, requiring ongoing dialogue among developers, regulators, local authorities and communities to weigh the five criteria and adapt as evidence emerges.
Gene drive deployment on islands: governance, risk assessment and community engagement
Islands present a distinct governance opportunity for EGD deployment. Their contained scale and high ecological or epidemiological stakes allow regulators, managers and communities to observe outcomes at a manageable level while generating evidence that can inform larger-scale decisions.
Policy frameworks are adapting to make this possible. Brodie et al. (2025) show that synthetically assisted conservation translocation and biocontrol align naturally with existing paradigms: classical translocation for species recovery and integrated pest management for invasive species. This alignment integrates EGD into familiar regulatory pathways rather than requiring entirely new categories. It also addresses gaps in instruments such as the Cartagena Protocol on Biosafety, which predate precise, self-limiting constructs. As I examined in Gene drive regulations: are we stuck in debate?, prolonged polarisation has slowed progress, but shifting toward proportionate, benefit-weighted assessments (explicitly weighing the severe cost of inaction against potential risks) opens a practical route forward. Islands, where the consequences of both action and inaction are sharply visible, offer ideal sites to test and refine such balanced governance in real conditions.
Stakeholder engagement emerges as the decisive factor. Lanzaro et al. (2021), Gierus et al. (2022) and Sudweeks et al. (2019) emphasise that early, continuous dialogue with communities, authorities and publics is essential for legitimacy and adaptive decision-making. On islands, small populations and strong environmental connections make this dialogue more direct and productive than in larger systems. Vorsino et al. (2025) illustrate how science-culture exchange and citizen-science initiatives in the Pacific can build lasting local expertise and infrastructure. In How community voices shape the future of gene drive research, I reviewed evidence from nearly 20 studies showing that meaningful inclusion (from information sharing to informed consent) strengthens social licence, enriches data quality, supports adaptive management and produces more equitable results. The same dynamic holds for malaria-endemic African islands: collaborative site evaluation, joint monitoring and shared decisions turn residents into knowledgeable partners who help shape outcomes.
These elements (policy reframing that embraces benefit-risk balance, and engagement that builds local ownership) create the conditions for responsible deployment. The literature does not demand flawless certainty but deliberate, iterative steps that capitalise on islands’ advantages. When governance and community processes align with technical capability, islands could demonstrate that EGD delivers measurable conservation and health gains. The path ahead seems clear: responsible deployment here could establish proof of concept, refine practice and build the foundation for broader, carefully scaled application.
Key takeaways
- Islands are proposed as initial sites for gene drive field trials due to their geographic boundaries
- Models suggest strong potential for suppression (e.g. up to 98% vector reduction), but outcomes depend on migration and spatial structure
- Complete containment is unlikely in most real-world island systems due to human and environmental connectivity
- Genomic approaches may enable partial localisation of gene drives
- Operational, social and governance factors are as critical as biological ones
Frequently asked questions (FAQs)
Why are islands considered for gene drive field trials?
Islands are considered promising settings for early gene drive field trials because their geographic boundaries may help limit dispersal of target species while allowing clear measurement of outcomes. Many islands are also relatively well-characterised ecologically and, in some cases, host fewer closely related species, which can simplify monitoring. These features make islands attractive environments to generate the first real-world evidence on how gene drives perform outside laboratory.
Can gene drives be contained on islands?
Islands may offer conditions that support a higher degree of containment compared to mainland settings, particularly where geographic and genetic isolation are strong. However, most island systems are not completely closed, as movement of organisms can still occur through human activities or environmental processes. Current research therefore focuses on how containment can be optimised and managed, rather than assumed to be absolute.
What risks exist for gene drive spread from islands to the mainland?
Research has explored how gene drives might move beyond target areas through natural dispersal or human-mediated transport. At the same time, a range of strategies are being developed to reduce this likelihood, including designs that limit spread or target locally specific genetic variants. Islands provide an opportunity to study these dynamics in real-world conditions and to evaluate how different approaches perform in practice.
Are island trials representative of mainland conditions?
Island trials can provide important early insights into gene drive performance, particularly in terms of spread, ecological interaction, and operational feasibility. At the same time, islands differ from mainland environments in factors such as connectivity, species diversity, and population structure. For this reason, island studies are generally seen as a first step in a broader evidence pathway, contributing valuable data that can inform future assessments in more complex settings.
What species are being targeted for gene drive use on islands?
Two main application areas are being explored. In conservation, gene drives are being investigated as a potential tool to control invasive rodents, such as mice and rats, which are major drivers of biodiversity loss on islands. In public health, research focuses on mosquitoes, particularly Anopheles species that transmit malaria. These use cases highlight how island trials could support both biodiversity protection and disease control efforts.
References
The literature map below reflects the set of recent studies included in this review, selected based on relevance to island conservation, modelling and genetics, and responsible deployment.
Birand, A., Cassey, P., Ross, J. V., Russell, J. C., Thomas, P., & Prowse, T. A. A. (2022) Gene drives for vertebrate pest control: realistic spatial modelling of eradication probabilities and times for island mouse populations. https://doi.org/10.1111/mec.16361
Brodie, J. F., Emmel, A., Wiedenheft, B., Sandler, R. L., Redford, K. H., Schultz, C. A., Moehrenschlager, A., Mark-Shadbolt, M., Kamau, W. S., Helm, J. E., Gendron, W. A. C., Dunn, S. L., & Schwartz, M. K. (2025) Synthetically assisted conservation and the application of emerging biological technologies for the protection of biodiversity. https://doi.org/10.1111/conl.13114
Connolly, J. B., Burt, A., Christophides, G., Diabate, A., Habtewold, T., Hancock, P. A., James, A. A., Kayondo, J. K., Lwetoijera, D. W., Manjurano, A., McKemey, A. R., Santos, M. R., Windbichler, N., & Randazzo, F. (2024) Considerations for first field trials of low threshold gene drive for malaria vector control. https://doi.org/10.1186/s12936-024-04952-9
Dhole, S., Vella, M. R., Lloyd, A. L., & Gould, F. (2018) Invasion and migration of spatially self-limiting gene drives: a comparative analysis. https://doi.org/10.1111/eva.12583
Gierus, L., Birand, A., Bunting, M. D., Godahewa, G. I., Piltz, S. G., Oh, K. P., Piaggio, A. J., Threadgill, D. W., Godwin, J., Edwards, O., Cassey, P., Ross, J. V., Prowse, T. A. A., & Thomas, P. Q. (2022) Leveraging a natural murine meiotic drive to suppress invasive populations. https://doi.org/10.1073/pnas.2213308119
Greenbaum, G., Feldman, M. W., Rosenberg, N. A., & Kim, J. (2021) Designing gene drives to limit spillover to non-target populations. https://doi.org/10.1371/journal.pgen.1009278
Hancock, P. A., North, A., Leach, A. W., Winskill, A., Ghani, A. C., Godfray, H. C. J., Burt, A., & Mumford, J. D. (2024) The potential of gene drives in malaria vector species to control malaria in African environments. https://doi.org/10.1038/s41467-024-53065-z
Hay, B. A., & Guo, M. (2022) Gene drive-mediated population elimination for biodiversity conservation: when you come to a fork in the road, take it. https://doi.org/10.1073/pnas.2213308119
IPBES (2019): Global assessment report on biodiversity and ecosystem services of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services. https://doi.org/10.5281/zenodo.3831673
Jones, R. T., Pretorius, E., Ant, T. H., Bradley, J., Last, A., & Logan, J. G. (2021) Use of islands and cluster-randomized trials to investigate vector control interventions: a case study on the Bijagós archipelago, Guinea-Bissau. https://doi.org/10.1098/rstb.2019.0807
Kim, J., Harris, K. D., Kim, I. K., Shemesh, S., Messer, P. W., & Greenbaum, G. (2023) Incorporating ecology into gene drive modelling. https://doi.org/10.1111/ele.14194
Lanzaro, G. C., Campos, M., Crepeau, M., Cornel, A., Estrada, A., Gripkey, H., Haddad, Z., Kormos, A., & Palomares, S. (2021) Selection of sites for field trials of genetically engineered mosquitoes with gene drive. https://doi.org/10.1111/eva.13283
Oh, K. P., Shiels, A. B., Shiels, L., Blondel, D. V., Campbell, K. J., Saah, J. R., Lloyd, A. L., Thomas, P. Q., Gould, F., Abdo, Z., Godwin, J. R., & Piaggio, A. J. (2021) Population genomics of invasive rodents on islands: genetic consequences of colonization and prospects for localized synthetic gene drive. https://doi.org/10.1111/eva.13210
Pan, M., & Champer, J. (2022) Making waves: comparative analysis of gene drive spread characteristics in a continuous space model. https://doi.org/10.1101/2022.11.01.514650
Sudweeks, J., Hollingsworth, B., Blondel, D. V., Campbell, K. J., Dhole, S., Eisemann, J. D., Edwards, O., Godwin, J., Howald, G. R., Oh, K. P., Piaggio, A. J., Prowse, T. A. A., Ross, J. V., Saah, J. R., Shiels, A. B., Thomas, P. Q., Threadgill, D. W., Vella, M. R., Gould, F., & Lloyd, A. L. (2019) Locally fixed alleles: a method to localize gene drive to island populations. https://doi.org/10.1038/s41598-019-51994-0
Vorsino, A. E., Harvey-Samuel, T., Hapairai, L. K., Bossin, H., & Bellinger, M. R. (2025) Genetic biocontrol strategy considerations for mosquito control in the Pacific Island Countries and Territories. https://doi.org/10.1016/j.cois.2025.101484
Willis, K., & Burt, A. (2019) Double drives and private alleles for localised population genetic control. https://doi.org/10.1371/journal.pgen.1009333