Stability versus specificity in polygenic disease resistance: an operational distinction in search of a statistical one

Question: How have multi-environment genome-wide association studies in crops partitioned polygenic disease resistance into stable versus environment-specific genomic regions, and what statistical frameworks (multi-trait GWAS, genotype-by-environment interaction models) are used to do so?

Answer

Across the retrieved literature, the stable/specific partition is overwhelmingly operational rather than modelled: authors run per-environment or per-isolate association scans and then define "stable" regions as the intersection of significant marker sets, or as loci surviving a threshold on estimated effect variance across environments, while genotype-by-environment (GxE) interaction is documented through reported interaction terms and heterogeneous evidence rather than estimated as a partitioned variance component for resistance. When formal multi-environment machinery is deployed, it mostly arrives from yield genetics (QxE linear mixed models, reaction-norm frameworks, environment kernels for prediction) and is rarely fitted to disease data; conversely, the axis along which "environment" varies in resistance studies is frequently pathogen genotype or temperature rather than the abiotic field environment, so the stable/specific contrast conflates race-nonspecificity with environmental robustness. The strongest empirical result that cuts against the standard framing is that cross-environment meta-analysis of a multi-site disease-resistance programme found no common fixed-effect association at 5% FDR while exposing extensive heterogeneity, separating compact haplotype-validation targets from background- and environment-dependent signals [4]; and where isolate × genotype interaction was explicitly tested, every mapped association was isolate-specific [24]. Taken together, these papers support a position in which "stable" loci in polygenic disease resistance are best understood as adult-plant/partial-resistance and multi-isolate-supported regions detected by replication and variance ranking, not as statistically demonstrated GxE-free effects.

"Stable" regions are currently defined by replication across environments and isolates, not by an interaction test

The dominant practice is to call a locus stable when it recurs across independent environments, isolates, or panels. In a 377-line wheat panel screened with 20 stripe rust isolates at the seedling stage and field races at the adult stage, 77 loci were mapped, of which nine were effective at the adult plant stage and did not overlap known adult-plant resistance genes [15]. In 297 Watkins landraces tested against six isolates spanning three pathogen lineages, 18 of 87 QTLs were supported by multiple isolates and 14 by multiple statistical models, and several accessions were stable across all lineages [23]. In barley bacterial leaf streak, ten accessions showed consistently low severity across four years and five marker-trait associations were recovered by at least two of four mixed-model implementations [21]. Wheat scald mapping likewise found one large chromosome 3H region recovered across all models [3]. What such designs establish is repeatability under the sampled environments; none of them tests whether the locus effect differs between environments.

Where environment specificity is visible, it is usually extreme rather than partial

Direct tests of interaction tend to show near-complete specificity of disease associations. In Septoria tritici blotch GWAS of 337 soft winter wheat lines inoculated with two isolates, isolate × genotype interaction was significant and all 12 mapped associations were isolate-specific and necrosis- or pycnidia-specific [24]. Temperature-resolved GWAS of bacterial wilt in tomato found no candidate genes shared between the 28 °C and 32 °C phenotyping conditions [6]. At the level of individual QTL, a stripe rust locus carried by the susceptible parent of a doubled-haploid population explained up to 30% of variation but was effective at a single Washington location [5]. These results argue that the environment-specific class is not a residual of noisy stable loci but a genuine, often large, class of effects.

Formal GxE frameworks exist but have mostly been built and tested outside disease resistance

The methodological core of GxE-aware association is a multi-environment linear mixed model that simultaneously fits main and interaction polygenic terms: a two-kinship-matrix model was more powerful for QTL-by-environment detection than meta-analysis or inclusive composite interval mapping, validated on rice traits across four environments and barley lodging across six [16]. Prediction-side developments include covariance structures derived from matrix multiplication of genotype and environment kernels, which improved accuracy up to 13.2% over Hadamard-based models across 11 wheat and rice datasets [7], reaction-norm models with weather covariates that were best in cross-validation schemes predicting incomplete trials but remained poor (0.10–0.20) when predicting new lines in new environments [9], and multi-trait/multi-environment mixed-model software enabling complex covariance structures [19]. None of these is applied to disease severity in the retrieved set, which is a substantial transfer gap.

Reaction-norm GWAS is the framework most exposed to being uninterpretable for plasticity

The most consequential caution concerns index-based reaction norms: replacing the environmental mean with a climate-derived index that correlates with environmental means at ρ caps the loading on plasticity-specific variation at √(1−ρ²), and simulated power to detect plasticity-specific loci falls from 0.64 at ρ = 0 to 0.009 at ρ = 0.99, so that the algorithm minimises the channel carrying the signal it is meant to detect [10]. The authors recommend reporting τ alongside ρ and reading slope loci below τ ≈ 0.15 as mean-performance loci [10]. A parallel argument dissolves the trait/environment boundary conceptually: GxE can be treated as a special case of pleiotropy, and the discrete null-hypothesis testing convention systematically underestimates pleiotropic interactions [13]. Absent such reporting, "environment-specific" loci in disease GWAS may be mean-performance loci mislabelled.

Effect-variance ranking and haplotype-level integration recover recurrent regions that single-environment scans miss

Ranking haploblocks by haplotype effect variance rather than by per-environment significance yielded 32 of the top 100 stripe rust haploblocks shared between an elite Australian panel and a landrace panel, and 29 comparable leaf rust blocks [2]. Haplotype-level analysis of net blotch in 950 barley accessions resolved 40 QTL, 15 shared between the two disease forms, with roughly 60% of haploblocks showing positive local genetic correlations and dual-disease haplotype stacking outperforming single-disease stacking [20]. Hotspot analysis of resistance to four soil-borne wheat pathogens found about two-thirds of hotspot components shared among pathosystems, but explicitly noted that evidence for excess sharing varied with the null model used, and that individual marker associations transferred poorly to an independent panel [1]. This is the clearest statement in the retrieved set that the stable fraction is scale-dependent — present at the level of regions and gene families, weak at the level of individual markers.

Multi-trait integration substitutes for, and sometimes conflates with, environmental generality

Multi-trait meta-QTL integration in maize classified hotspots into multi-trait hubs, pathway-focused clusters, and major-effect loci, and separately classified meta-QTLs by environmental support, with 42% supported predominantly by optimal-condition QTLs and 28% by stress-condition QTLs [11]. Cross-trait grouping in grapevine collapsed 270 significant marker tests into 54 candidate multi-trait regions, but the authors state this did not establish biological pleiotropy [4]. Because GxE is formally a pleiotropy problem [13], multi-trait grouping is not a substitute for an interaction test.

Conflicts and gaps

The central conflict is quantitative: one programme found zero common fixed-effect associations across environments [4] while others report substantial sharing — two-thirds of hotspot components [1], 18 of 87 multi-isolate-supported QTLs [23], 32 of 100 top haploblocks [2]. These are not commensurable because "sharing" is defined by different null models, thresholds, and marker units. A second gap is that ordinal disease scores violate the Gaussian assumptions of the mixed models used: whole-genome ordered probit regression has been proposed to deconstruct ordinal ratings into marker effects and produce genome-informed categorical probabilities independent of location-, rater- and plot-specific noise [17], but it has not been embedded in a multi-environment interaction framework. Third, stability parameters are used at the parental or hybrid level rather than as GWAS phenotypes: GGE biplots were used to confirm tester discrimination and representativeness in maize lethal necrosis testcrosses with significant GxE [8], but no retrieved study maps loci for a stability statistic. Fourth, plasticity-QTL mapping in a pathogen shows the same unresolved ambiguity as in hosts — 75% of plasticity QTL overlapped mean QTL, implying pleiotropy, while 25% mapped elsewhere, implying epistasis [14].

What would change the answer

Fitting the two-kinship QxE model [16] to a disease panel with at least ten environments and reporting the interaction variance component per locus would convert the stable/specific dichotomy into an estimated quantity. Orthogonal confirmation requires near-isogenic lines carrying individual and stacked resistance QTL across locations and years, as done for pea Aphanomyces root rot where major-effect QTL effects were confirmed across environments and genetic backgrounds [12]. Reaction-norm disease GWAS should report τ alongside ρ before any slope locus is called plasticity-specific [10], and allele effects should be modelled as functions of measured environmental variables with mediation analysis, as done for vernalization and photoperiod alleles across 219 site-years [22]. Pairing ordinal-probit genome-informed scores [17] with isolate panels that permit explicit GxI terms [24] would separate pathogen-driven from climate-driven specificity. Finally, structural variants associated with GxE but in linkage disequilibrium with nearby SNPs [18] should be tested directly rather than assumed captured.

References

[1] Pariyar, S., Dadshani, S., Dababat, A. A., Erginbas-Orakci, G., Habash, S., Wen, N., Grundler, F. M. W., Leon, J.. "Integrated genomic analyses reveal hierarchical organisation of quantitative resistance to biologically distinct soil-borne pathogens in wheat". bioRxiv. 2026-10-19T00:00:00.000Z. URL: https://doi.org/10.64898/2026.10.14.758312

[2] Yadav, S., Dillon, S., McNeil, M., Dinglasan, E., Jambuthenne, D., Mago, R., Dodds, P. N., Hickey, L., Hayes, B.. "Advancing rust resistance in elite wheat with haplotype mapping and a novel introgression strategy". bioRxiv. 2025-09-25T00:00:00.000Z. URL: https://doi.org/10.1101/2025.09.22.677963

[3] Kolkman, J. M., Sepp, S. S., Kunze, K. H., Bergstrom, G. C., Sorrells, M. E.. "Genome-wide association analysis of resistance to scald in an adapted multiparent winter malting barley population". bioRxiv. 2026-04-11T00:00:00.000Z. URL: https://doi.org/10.64898/2026.03.12.711358

[4] Borrelli, C., Prado, E., Dumas, V., Arnold, G., Onimus, C., Butterlin, G., Jaegli, N., Wiedemann-Merdinoglu, S., Lacombe, M.-C., Dorne, M.-A., Umar-Faruk, A., Chaumonnot, S., Valentin, S., Ley, L., Reynard, J.-S., Spring, J.-L., Duchene, E., Schneider, C., Merdinoglu, D., Avia, K.. "Uncovering the genetic basis of agronomic traits in over 1,000 grapevine genotypes derived from a disease resistance breeding program". bioRxiv. 2026-08-18T00:00:00.000Z. URL: https://doi.org/10.1101/2025.10.05.680539

[5] Sharma, R., Wang, M., Chen, X., Carver, B. F., Guttieri, M., St. Amand, P., Bernardo, A., Bai, G., Ara, A. M., Jordan, K. W., Aoun, M.. "Identification of stripe rust adult plant resistance genes in the hard winter wheat cultivar Bakers Ann". bioRxiv. 2025-09-07T00:00:00.000Z. URL: https://doi.org/10.1101/2025.09.04.674321

[6] Belny, A., Desaint, H., Rigal, M., Carrere, S., Gouzy, J., Bonnet, G., Labourey, C., Albert, E., Grivet, L., Roux, F., Vailleau, F., Mary-Huard, T., Berthome, R.. "New resistance to bacterial wilt in heat-stressed tomato is revealed by two-reference Genome Wide Association". bioRxiv. 2025-07-31T00:00:00.000Z. URL: https://doi.org/10.1101/2025.07.28.667131

[7] Montesinos-Lopez, O. A., Montesinos-Lopez, A., Montesinos-Lopez, J. C., Crossa, J., Dreisigacker, S., Hernandez-Suarez, C. M., Ortiz, R.. "A novel matrix multiplication framework for modeling genotype-by-environment interaction in genomic prediction". bioRxiv. 2026-05-15T00:00:00.000Z. URL: https://doi.org/10.64898/2026.05.11.724414

[8] Gowda, M., Beyene, Y., L.M., S., Ogugo, V., Amadu, M. K., Chaikam, V.. "Genetic Insights from Line x Tester Analysis of Maize Lethal Necrosis Testcrosses for Developing Multi-Stress-Resilient Hybrids in Sub-Saharan Africa". bioRxiv. 2025-12-09T00:00:00.000Z. URL: https://doi.org/10.64898/2025.12.07.692857

[9] Acharya, S. R., Garcia-Abadillo, J., Lyerly, J., Brown-Guedira, G., Jarquin, D., Bandillo, N.. "Reaction Norm Modeling of High-Dimensional Genomic and Environmental Data Improves Prediction Accuracy in Winter Wheat". bioRxiv. 2026-05-08T00:00:00.000Z. URL: https://doi.org/10.64898/2026.05.05.722758

[10] Jenkins, S. D., Graef, G. L.. "When Does Reaction Norm GWAS Discover Plasticity? The Residual Channel in Environmental Index-Based Genetic Dissection". bioRxiv. 2026-07-28T00:00:00.000Z. URL: https://doi.org/10.64898/2026.07.27.741005

[11] Parthasarathy, S., Koehler, K., Rocheford, T.. "Multi-Trait Meta-QTL Analysis Reveals Genomic Hotspot Classes for Strategic Maize Improvement". bioRxiv. 2026-08-21T00:00:00.000Z. URL: https://doi.org/10.64898/2026.06.06.730627

[12] LAVAUD, C., RIVIERE, J.-P., VETEL, P., AUBERT, G., MARGET, P., ROULLET, G., HERBOMMEZ, J.-F., DECLERCK, P., Pilet-Nayel, M.-L.. "QTL combinations associated with field partial resistance to aphanomyces root rot in pea Near-Isogenic Lines". bioRxiv. 2025-03-02T00:00:00.000Z. URL: https://doi.org/10.1101/2025.02.28.640773

[13] Ellis, T. J.. "How and why to quantify pairwise pleiotropy and genotype-by-environment interactions". bioRxiv. 2026-01-12T00:00:00.000Z. URL: https://doi.org/10.64898/2026.01.09.698568

[14] Stapley, J., McDonald, B. A.. "Genetic mapping of trait plasticity in a plant pathogenic fungus reveals genetic architecture and candidate genes for plasticity". bioRxiv. 2026-08-22T00:00:00.000Z. URL: https://doi.org/10.64898/2026.08.18.745209

[15] Sahin, B., Wang, M., Zurn, J. D., Xu, X., Bai, G., Akhunova, A., Chen, X., Akhunov, E.. "Genetic architecture and evolution of stripe rust resistance uncovered using diverse panels of wheat lines and North American Puccinia striiformis f. sp. tritici isolates". bioRxiv. 2025-12-11T00:00:00.000Z. URL: https://doi.org/10.64898/2025.12.10.693471

[16] Zhao, F., Wang, L., Xu, S.. "Identification of QTL-by-Environment Interaction by Controlling Polygenic Background Effect". bioRxiv. 2024-12-29T00:00:00.000Z. URL: https://doi.org/10.1101/2024.12.29.630685

[17] Check, J. C., Technow, F., Totir, L. R.. "An industry perspective on whole genome-informed hybrid maize disease resistance characterization to improve breeding decisions". bioRxiv. 2026-08-05T00:00:00.000Z. URL: https://doi.org/10.64898/2026.07.30.741863

[18] Catlin, N. S., Agha, H. I., Platts, A. E., Munasinghe, M., Hirsch, C. S., Josephs, E. B.. "Structural variants contribute to phenotypic variation in maize". bioRxiv. 2024-12-03T00:00:00.000Z. URL: https://doi.org/10.1101/2024.06.14.599082

[19] COVARRUBIAS PAZARAN, G. E.. "lme4breeding: a fast linear mixed model for multi-trait/multi-environment experiments". bioRxiv. 2025-09-29T00:00:00.000Z. URL: https://doi.org/10.1101/2024.05.17.594795

[20] Liu, D., Zhang, X., Snyman, L., Garrard, T., Wallwork, H., Dadu, H., Maclean, M., Tong, J., Chen, C., Gamaralalage, D. J., Periyannan, S., Hickey, L., Hayes, B. J., Dinglasan, E.. "Haplotype-based insights into the genetic architecture of net blotch resistance in barley". bioRxiv. 2026-08-09T00:00:00.000Z. URL: https://doi.org/10.64898/2026.07.23.740211

[21] Velasco, D. D. P., Shi, G., Brueggeman, R. S., Horsley, R. D., Liu, Z., Baldwin, T. T.. "A subset of world barley collection was used in the identification of sources of resistance and marker-trait association for resistance for bacterial leaf streak". bioRxiv. 2025-11-03T00:00:00.000Z. URL: https://doi.org/10.1101/2025.10.31.685662

[22] DeWitt, N., Winn, Z., Guedira, M., Lyerly, J., Maltecca, C., Ward, B., Stafstrom, W., Babar, M. A., Bandillo, N., Boyles, R., Harrison, S., Howell, K., Mason, R. E., Melson, E., Mergoum, M., Mohammadi, M., Murphy, J. P., Olson, E., Rutkoski, J., Santantonio, N., Smith, J. H., Sneller, C., Sorrells, M., Sutton, R., Tiwari, V., Van Sanford, D. A., Brown-Guedira, G.. "Environmentally-mediated yield effects of vernalization and photoperiod alleles in historic winter wheat trials". bioRxiv. 2026-09-24T00:00:00.000Z. URL: https://doi.org/10.64898/2026.09.18.752491

[23] Singh, J., Awan, M. J. A., Kumar, N., Holden, S., Khangura, R. S., Singh Brar, G.. "Watkins wheat landraces: a treasure of stripe rust resistance alleles identified using multi-model association analyses". bioRxiv. 2026-03-13T00:00:00.000Z. URL: https://doi.org/10.64898/2026.03.11.711137

[24] Aouini, L., Gaire, R., Scofield, S., Brown-Guedira, G., Mohammadi, M., Goodwin, S. B.. "Genome-Wide Association study in a US soft winter wheat population reveals novel and known sources of resistance to the Septoria tritici blotch pathogen Zymoseptoria tritici". bioRxiv. 2026-07-16T00:00:00.000Z. URL: https://doi.org/10.1101/2025.09.30.679614

How this survey was made

This survey answers one question: “How have multi-environment genome-wide association studies in crops partitioned polygenic disease resistance into stable versus environment-specific genomic regions, and what statistical frameworks (e.g., multi-trait GWAS, genotype-by-environment interaction models) are used to do so?”

We keep our own collection of recent preprints and refresh it every day from bioRxiv, medRxiv, arXiv, PsyArXiv, SocArXiv and ChemRxiv. For each survey, we search that collection for recent work and keep the strongest matches: papers that use the same terms as the question, and papers on the same topic in different words.

The text was written from the titles and abstracts of those papers; the full texts were not read. Statements that draw on a particular paper are marked with a number in brackets. The reference list just above is the 24 papers whose work this text draws on, chosen from the 150 papers those searches returned; the rest of the shortlist is not printed. It is not a complete review of everything published on the question.

Treat this as a starting point and a reading list, not a peer-reviewed review. The searches behind this survey ran on October 22, 2026 and covered work from the previous two years.