In a groundbreaking advancement in forest genetics, researchers have unveiled novel genomic-based causal models poised to revolutionize the way breeding values (BVs) are estimated in tree populations. Traditionally reliant on pedigree-based individual-tree mixed models, known as ABLUP, forest genetic evaluations have now crossed a pivotal threshold by integrating the precision and depth of genomic data through innovative Markov causal frameworks. This fusion of classical genetics with high-density genomic markers heralds a new era in tree breeding, promising both enhanced accuracy and computational efficiency.
The ABLUP methodology, a long-standing cornerstone in forest genetic evaluations, operates on a Markovian causal assumption: the breeding value of any individual tree can be represented as a linear function of its parental BVs. Here, the regression coefficients—signifying relationships such as parent to offspring—are fixed at one-half, reflecting the expected genetic contribution from each parent under Mendelian inheritance principles. While this approach has provided substantial insights across numerous breeding programs, its reliance on genealogical data constrains its ability to capture the nuanced genetic diversity present within and across families.
Motivated by the explosive growth in genomic technologies, the current study moves beyond pedigree constraints by developing two new causal models that incorporate coefficients derived from genomic relationship matrices. These matrices, constructed from thousands of single nucleotide polymorphism (SNP) markers distributed throughout the genome, offer a more detailed and individualized map of genetic relatedness. By replacing the fixed coefficients of traditional models with data-driven genomic analogs, the researchers aimed to model the inheritance structure with unprecedented specificity, capturing intricate patterns of genetic variance that pedigrees alone cannot resolve.
This advance was put to the test using a remarkable dataset comprising a four-generation population of Eucalyptus grandis — a species esteemed for its rapid growth and economic value in forestry. The population consisted of over 3,000 genotyped individuals, each characterized by more than 14,000 SNP markers, providing a rich genomic landscape. Within this cohort, six distinct phenotypic traits were measured in roughly 1,200 trees across the first three breeding cycles, enabling a robust phenotypic-genotypic association framework to assess the performance of these emerging causal models.
When benchmarked against traditional pedigree-based ABLUP models and the widely employed genomic best linear unbiased prediction (GBLUP) models, the genomic causal models distinguished themselves notably. Estimates for trait heritability— the proportion of observed variation attributable to genetic factors—were markedly higher in both the pedigree- and genomic-based causal frameworks compared to GBLUP. This suggests that the causal models more effectively disentangle the genetic signal from the noise, a critical facet in breeding value estimation that directly influences selection decisions.
Furthermore, genetic mean estimates, which provide an average breeding value across the population or subgroups, also trended higher under the causal approaches, reinforcing the notion that these models may facilitate improved detection of favorable genetic effects. Such refined estimates could substantially steer breeding programs toward more effective parent selection, ultimately accelerating genetic gain and trait improvement over successive generations.
Interestingly, realized genetic gains—actual improvements observed through selection—remained comparable across all models. However, the predictions of genetic gains derived from the causal models aligned closer with these realized results than those obtained from GBLUP. This enhanced concordance underscores a crucial advantage for causal models: their superior fidelity in forecasting breeding outcomes, a vital feature for reliable long-term breeding strategies.
Concurrently, GBLUP, despite its somewhat lower precision in matching realized gains, retained an edge in predictive performance. That is, it offered better prediction accuracy on unseen data, albeit with wider confidence intervals. This contrasting performance profile illuminates a trade-off inherent in current genomic prediction paradigms—between precision and predictive robustness—highlighting the complementary roles that different models might play depending on program goals.
Beyond predictive qualities, the computational load associated with genomic evaluations often poses practical challenges, especially as datasets balloon in size and complexity. Notably, the causal models developed within this research demonstrated a lower computational burden relative to GBLUP, making them appealing candidates for deployment in large-scale breeding operations. This balance of precision, predictive quality, and computational efficiency situates these new models as promising tools for modern forestry programs.
One transformative feature of the causal models lies in their ability to resolve intra-familial variation — distinctions in breeding values among siblings arising from the random assortment of alleles during meiosis. Conventional pedigree models treat siblings identically, given their equal expected relatedness, glossing over this layer of genetic diversity. By integrating genomic data into causal coefficients, the new models capture these subtle yet critical differences, potentially unlocking selection potential within families that would otherwise remain hidden.
This aspect holds particular significance for forest tree breeding, where balancing genetic gain against the maintenance of diversity is paramount for sustaining forest health under shifting environmental conditions. More accurate and granular estimates of breeding values empower breeders to craft selection strategies that are both effective and genetically diverse, thus fostering resilience in future forests.
The study’s use of Eucalyptus grandis, with its economic and ecological importance, further enhances the relevance and applicability of the findings. Fast growth, adaptability, and wood quality traits in Eucalyptus are of great interest to the forestry industry worldwide, making precise breeding models indispensable for optimizing production and sustainability.
In sum, this pioneering work presents the first genomic Markov causal models tailored for forest tree breeding, marking a significant leap forward from pedigree-based approaches. By harmonizing classical genetic theory with contemporary genomic data, the researchers offer a compelling vision for future breeding endeavors — one where detailed genetic insight, computational feasibility, and predictive reliability converge to accelerate genetic gain.
Looking ahead, the integration of such models into breeding pipelines promises to refine genetic evaluations, enabling the forestry sector to meet escalating demands for timber, bioenergy, and ecosystem services in a changing climate. As genomic technologies become more accessible and datasets grow ever richer, causal genomic models stand poised to become foundational tools in sustainable, precision forest breeding.
This innovative framework also opens avenues for further research, including the incorporation of environmental variables, epistatic interactions, and the extension to multi-species breeding programs. The evolution of causal inference within genomics thus not only enhances current methodologies but also lays groundwork for a deeper understanding of genetic architecture influencing complex traits.
Ultimately, the adoption of genomic Markov causal models reflects a broader shift within applied genetics — moving from fixed, simplified assumptions towards dynamic, data-driven representations of heredity. This transition unlocks the potential for more nuanced, personalized genetic predictions, thereby transforming breeding from an art into a precision science.
Such advances align closely with global imperatives to sustainably manage natural resources, combat climate change, and secure food and fiber supplies. By enhancing selection accuracy and accelerating breeding cycles, genomic causal models contribute to more resilient, productive forestry systems capable of thriving in the 21st century and beyond.
As forest geneticists and breeders worldwide explore and refine these new tools, the promise of blending genomic intricacy with causal clarity stands as a testament to the power of interdisciplinary innovation to revolutionize how we steward the natural world.
Subject of Research: Forest genetic evaluations and breeding value prediction using genomic Markov causal models in Eucalyptus grandis.
Article Title: Forest tree breeding using genomic Markov causal models: a new approach to genomic tree breeding improvement.
Article References:
Jurcic, E.J., Dutour, J., Villalba, P.V. et al. Forest tree breeding using genomic Markov causal models: a new approach to genomic tree breeding improvement. Heredity (2025). https://doi.org/10.1038/s41437-025-00755-z
Image Credits: AI Generated
