A 2023 large-scale genomic analysis of 2,839 rice hybrid cultivars and 9,839 segregating individuals (Gu et al., Nature Genetics) showed that genetic complementation at dominance-effect loci contributes more to indica–japonica intersubspecific yield heterosis than overdominance-effect loci, yet left two coupled gaps explicitly acknowledged by the authors: loci scored as overdominant may instead be pseudo-overdominance from tightly linked variants in opposite phase, with no genome-scale way to tell them apart; and adding dominance terms to genomic-selection models did not improve prediction accuracy, because locus-specific dominance and epistatic effects cannot yet be estimated reliably. Later work through 2026 — new heterotic-QTL scanning methods in wheat, long-read haplotype analysis in pearl millet that still cannot separate overdominance from pseudo-overdominance, multi-omics rice studies, and reviews stating that the molecular basis of heterosis remains unresolved — has not closed either gap. The open problem is to resolve the dominance/overdominance architecture of rice heterosis at the level of individual loci and to demonstrate that this resolved architecture improves prediction of hybrid performance.

Background

Heterosis (hybrid vigor) is the superior performance of the first-generation (F1F_1) offspring of two genetically distinct inbred parents relative to those parents. Three classical, non-exclusive explanations are debated: the dominance hypothesis (heterosis arises because deleterious recessive alleles inherited from one parent are complemented by favorable alleles from the other, so the hybrid masks inferior alleles), the overdominance hypothesis (the heterozygous allele combination at a single locus is intrinsically superior to either homozygote), and epistasis (interactions between alleles at different loci). A long-standing confound is pseudo-overdominance: two tightly linked loci carrying deleterious recessive alleles in opposite phase (each parent's deleterious allele at one locus sits next to the other parent's favorable allele at the neighboring locus) imitate single-locus overdominance in an F1F_1, because recombination almost never separates them. Distinguishing true from pseudo-overdominance matters practically: complementation and pseudo-overdominance are in principle fixable in an inbred line by recombination or genome editing, whereas true overdominance is not. Cultivated rice (Oryza sativa) has two main subspecies, indica and japonica; hybrids between them (intersubspecific hybrids) show strong heterosis but are difficult to breed. Gu et al. (2023) resequenced and phenotyped 2,839 hybrid rice cultivars and 9,839 segregating individuals, performed genome-wide association studies (GWAS, which statistically associate genotype variation at molecular markers with trait variation across individuals) to identify quantitative trait nucleotides (QTNs, the individual DNA-base variants contributing to quantitative traits), and found that dominance-effect loci explain more phenotypic variance in intersubspecific yield heterosis than overdominance-effect loci. They also built a genomic-selection model (a statistical model that predicts a cross's merit from genome-wide marker genotypes as a genomic estimated breeding value, GEBV). Two limitations they explicitly acknowledge define the open problem. First, they caution that their over-dominance loci 'may also include the pseudo-overdominance loci, which was conferred by tightly linked genes in opposite phase' — the classification is confounded at genome scale. Second, an additive-plus-dominance genomic-selection model did not improve prediction accuracy over an additive-only model (their Extended Data Fig. 9a); they attribute this to dominance effects varying in magnitude and sign across loci and to the added model complexity, concluding that 'current models were inadequate in handling the non-additive effect'. These two gaps are two faces of one issue: without locus-level resolution of phase and linkage, per-locus non-additive effects cannot be estimated well enough either to classify the mechanism or to improve prediction. Single-locus overdominance has been demonstrated in tomato and sorghum, and complementation in Arabidopsis, maize and rice, but a genome-scale, locus-resolved, and predictive account of heterosis in any major cereal remains lacking.

Problem Statement

For rice heterosis — in particular indica–japonica intersubspecific heterosis for yield-related traits — resolve the non-additive genetic architecture at the level of individual loci and make that resolution predictive. Concretely, Gu et al. (2023) leave open: (i) at genome scale, determine which loci currently classified as overdominant reflect true single-locus overdominance and which instead reflect pseudo-overdominance from tightly linked variants in opposite phase (including identification of the linked variants and their phase); and (ii) construct genomic estimation/prediction frameworks in which locus-specific dominance and epistatic effects are estimated reliably enough that incorporating them measurably improves prediction of hybrid performance relative to additive-only genomic selection — something Gu et al. (2023) found a pooled dominance term failed to do. These two elements form one scientific objective — a locus-resolved, predictive account of rice heterosis — and neither stands alone: the resolved architecture is predictive only when incorporating the locus-specific non-additive effects yields a measurable improvement in out-of-sample prediction of hybrid performance — on held-out crosses or by cross-validation, not by in-sample fit — over an additive-only genomic-selection baseline.

References

  1. Zhoulin Gu, Junyi Gong, Zhou Zhu, Zhen Li, Qi Feng, Changsheng Wang, Yan Zhao, Qilin Zhan, Congcong Zhou, Ahong Wang, Tao Huang, Lei Zhang, Qilin Tian, Danlin Fan, Yiqi Lu, Qiang Zhao, Xuehui Huang, Shihua Yang, Bin Han (2023), 'Structure and function of rice hybrid genomes reveal genetic basis and optimal performance of heterosis', Nature Genetics 55(10):1745-1756, DOI: 10.1038/s41588-023-01495-8, https://doi.org/10.1038/s41588-023-01495-8
  2. Zhoulin Gu, Bin Han (2024), 'Unlocking the mystery of heterosis opens the era of intelligent rice breeding', Plant Physiology 196(2):735-744, DOI: 10.1093/plphys/kiae385, https://doi.org/10.1093/plphys/kiae385
  3. Frank Hochholdinger, Peng Yu (2025), 'Molecular concepts to explain heterosis in crops', Trends in Plant Science 30(1):95-104, DOI: 10.1016/j.tplants.2024.07.018, https://doi.org/10.1016/j.tplants.2024.07.018
  4. Guoliang Li, Renate H. Schmidt, Yusheng Zhao, Jochen C. Reif, Yong Jiang (2025), 'Powerful one-dimensional scan to detect heterotic quantitative trait loci', Nature Communications 16:9697, DOI: 10.1038/s41467-025-65563-9, https://doi.org/10.1038/s41467-025-65563-9
  5. Marine Salson, Maud Duranton, Stella Huynh, Cédric Mariac, Christine Tranchant-Dubreuil, Julie Orjuela, Philippe Cubry, Anne-Céline Thuillet, Concetta Burgarella, Miguel de Navascués, Leïla Zekraouï, Marie Couderc, Sandrine Arribat, Nathalie Rodde, Adeline Barnaud, Adama Faye, Ndjido Kane, Yves Vigouroux, Cécile Berthouly-Salazar (2025), 'Interplay between large low-recombining regions and pseudo-overdominance in a plant genome', Nature Communications 16:6458, DOI: 10.1038/s41467-025-61529-z, https://doi.org/10.1038/s41467-025-61529-z
  6. Zelin Chen, Peiwen Zhu, Deyan Kong, Lijun Luo, Xinqiao Yu, Guolan Liu (2026), 'Decoding the genetic mechanism of heterosis in rice: insights from multi-omics and systems approaches', Annals of Botany 138(1):7-19, DOI: 10.1093/aob/mcag010, https://doi.org/10.1093/aob/mcag010
  7. Zhiwu Dan, Yunping Chen, Wenchao Huang (2025), 'Quantitative Insights into the Genetic Mechanisms of Crop Heterosis', Advanced Genetics 6(4):e00030, DOI: 10.1002/ggn2.202500030, https://doi.org/10.1002/ggn2.202500030
  8. Zhiwu Dan, Yunping Chen, Wenchao Huang (2025), 'Structural Variations Contribute to Subspeciation and Yield Heterosis in Rice', Plant Biotechnology Journal, DOI: 10.1111/pbi.70484, https://doi.org/10.1111/pbi.70484
  9. Zhiwu Dan, Yunping Chen, Wenchao Huang (2026), 'Machine learning-based prediction of dynamic heterosis for plant height with pathway biomarkers in rice', Theoretical and Applied Genetics 139(4):113, DOI: 10.1007/s00122-026-05218-5, https://doi.org/10.1007/s00122-026-05218-5
  10. Miaomiao Zhao, Xuelian Han, Ao Zheng, Yuwen Yan, Hanqing Zhang, Xiaoshuang Ma, Jingtian Wang, Yuanming Zhang (2026), 'Dissecting the genetic foundation of heterosis for rice complex traits using GWAS, TWAS and mGWAS', Journal of Integrative Agriculture, DOI: 10.1016/j.jia.2026.06.016, https://doi.org/10.1016/j.jia.2026.06.016
  11. Yiyong Zhao, Tao Li, Daliang Liu, Hao Yin, Liang Wang, Song Lu, Houlin Yu, Xinhao Sun, Taikui Zhang, Quanzhi Zhao (2025), 'The origin and evolution of cultivated rice and genomic signatures of heterosis for yield traits in super-hybrid rice', BMC Biology 23:153, DOI: 10.1186/s12915-025-02255-2, https://doi.org/10.1186/s12915-025-02255-2
  12. Nia Manlulu, Rogemae Ravela, Frodie Waing, Leonilo Gramaje (2025), 'Molecular and physiological basis of heterosis in hybrid rice performance', Molecular Breeding 45(5):49, DOI: 10.1007/s11032-025-01577-x, https://doi.org/10.1007/s11032-025-01577-x