Tectona grandis (teak) is a tree species highly appreciated for its high-quality hardwood and versatile industrial applications. Despite clonal teak plantations showing several advantages, such as reduced rotation time and increased wood volume production, the genetic uniformity of plantations makes them more susceptible to diseases. Ceratocystis wilt caused by Ceratocystis manginecans has been the most prevalent disease on teak plantations, reducing tree growth rate, wood quality and value. Planting resistant genotypes is the most effective measure to mitigate losses caused by the disease. Thus, this work aimed to assess the resistance to Ceratocystis wilt of 12 teak clones commercially planted in Brazil, as well as the inheritance of resistance in teak open-pollinated families. Five teak clones exhibited resistance to Ceratocystis wilt, and all investigated open-pollinated teak families segregated resistance. The findings support the interpretation that resistance to Ceratocystis wilt in teak is a quantitative trait with additive gene effects in determining this trait. The information obtained in this work is an important contribution to directing the efforts of teak breeding programs in the attempt to reduce the losses caused by Ceratocystis wilt.

High-throughput phenotyping using unmanned aerial vehicles (UAVs) and spectral vegetation indices has been proposed to overcome the cost and logistical constraints of manual measurements in multi-environment breeding trials. However, the reliability of models trained on spectral data to predict structural traits across genotypes and environments remains unclear. This study aimed to develop an approach for predicting soybean plant height (PH) and first pod insertion height (FPIH) using UAV-based vegetation indices acquired at the flowering stage, as well as to compare extreme gradient boosting (XGBoost), multilayer perceptron (MLP), random forest (RF), and multiple linear regression (MLR) models under realistic cross-validation scenarios. Trials were conducted across multiple seasons using UAV multispectral imagery, with PH and FPIH manually measured. The models were evaluated under five phenotyping scenarios: baseline calibration using all data; prediction in a completely unmeasured future season; estimation of missing genotypes within a partially sampled season; calibration using a small fraction of data from a new season; and prediction under absence of field records for specific genotypes across environments. When all data were used for calibration, non-linear models showed a high apparent accuracy. However, prediction in unseen seasons failed for all models, reflecting strong genotype & times; environment interactions. Under reduced phenotyping within the same environment network, the models maintained a robust accuracy for PH, whereas FPIH predictions declined to moderate levels. UAVbased models are reliable for interpolation, but limited for extrapolation without local calibration, enabling reductions of up to 80 % in manual measurements for PH and 20-30 % for FPIH.

Genotype-by-environment interaction (GEI) complicates variety recommendations for common bean ( Phaseolus vulgaris L.) in regions with high agroclimatic variability. This study aimed to dissect GEI in Brazilian multi-environment trials (METs) using a linear mixed models with factor analytic (FA) variance structures for the genotype-by-environment effects (FA mixed models) to identify high-yielding, stable genotypes and to characterize the environmental drivers underlying this interaction. Eleven genotypes were evaluated across 32 environments (combinations of location, year, and season). The FA model with four factors (FA4) efficiently captured 81.9% of the GEI variance, showing high accuracy (0.93) and generalized heritability (0.88). Using Factor Analytic Selection Tools (FAST), the model outputs were synthesized into metrics for overall performance (OP) and stability (RMSD). Genotypes CNFP16379 and CNFP16404 were identified as superior candidates for broad adaptation, combining high OP with small RMSD. In contrast, genotypes like CNFP16830 and IPR UIRAPURU, with high OP but low stability, were considered suitable for specific environments. To interpret the environmental basis of GEI, correlations were performed between the factor loadings and climatic covariates, revealing that the interaction was primarily driven by gradients related to water balance (precipitation, humidity, vapor-pressure deficit) and solar radiation. The results demonstrate that the FAST framework is a powerful tool for common bean breeding, enabling data-driven selection for both broad and specific adaptation while providing valuable insights into the environmental factors that modulate genotype performance. This integrated strategy enhances breeding efficiency and supports more precise cultivar deployment across diverse agroecosystems.

This research aims to investigate the patterns of inbreeding depression in an $S_{0:1}$ population of Eucalyptus spp., by examining autozygosity and genetic parameters, and studying the impact of an unequal number of selfed and crossed individuals within families on inbreeding depression estimations. Inbreeding has been less investigated in forest species, including eucalyptus, than in annual crops, largely because of their extended reproductive cycles, substantial genetic load, and the practical difficulties of performing controlled selfing. In this study, we self-pollinated 20 elite Eucalyptus spp. genotypes. From each self-pollinated genotype, 30 seeds were collected, resulting in seedlings that were subsequently planted in a field trial. A total of 600 individuals were established in a randomized complete block design trial and evaluated for growth traits at 3 years of age. Both the progenies and the 20 parent were genotyped using Single Nucleotide Polymorphisms chips. Inbreeding depression was evident, as indicated by a decline in diameter at breast height from 13.88 to 9.73 cm in selfed individuals compared with crossed ones. Moreover, the autozygosity observed in the most inbred individuals was primarily due to recent inbreeding, whereas in the best-performing individuals, it mainly resulted from ancient inbreeding events. Simulations highlighted that unbalanced sample sizes of selfed and crossed individuals within families could bias estimates of inbreeding depression. By integrating genomic data and advanced quantitative methods, this study provides new insights into the genetic consequences of self-pollinating eucalyptus, offering a foundation for managing inbreeding and enhancing genetic gains in perennial crops.

The macauba palm, Acrocomia aculeta, stands out as an optional resource in the vegetable oil production chain, meeting the global demand for this raw material. Still, since the macauba is a marginal species, the development of superior genetic materials can ensure uniformity in large-scale orchards and productions, which are prerequisites for the establishment of new crops in the agricultural scenario. The present work evaluated a macauba progeny test for the selection of promising genotypes with agronomic traits of interest. Thirty-six genetic accessions from half-sib families were evaluated. They were acquired from natural populations growing in two regions in Minas Gerais State, Brazil. The genotypic values were obtained based on vegetative, reproductive, and biometric data from the fruits, via mixed models, followed by genetic diversity analysis. Genetic parameter estimates were obtained using the Restricted Maximum Likelihood procedure from interactions in the mixed model equations. The study revealed high values of environmental variation between plots and varying magnitudes for CV_gi , CV_e and CV_r . The progenies, either isolated or as single population sets, revealed great genetic diversity. When seeking to develop cultivars with a lower number of crossings, the progenies from the Santa Luzia-MG region proved to be suitable for initiating crossings as they presented greater genetic uniformity. However, considering the diversity and genetic value of the Luz-MG families, the following crossings are recommended: TP6 × TP8, TP6 × TP18, TP6 × TP13, TP8 × TP18 and TP13 × TP18, as they have a better content of alleles favorable to genetic improvement.

Forest ecosystems play an essential role in maintaining biodiversity and mitigating global warming. Intensively managed forest plantations represent a major raw material source for pulp, timber and paper industries. The objective of this study was to apply the model identity test to compare the regression equations used in wood volume estimates of forest species. Five eucalypt species (E. camaldulensis, E. urophylla, E. saligna, E. grandis and E. urograndis) and Corymbia citriodora were evaluated in an experimental area of the Federal University of Mato Grosso do Sul. Five trees in each plot were measured from 56 to 71 months of age to assess the diameter at breast height (DBH) and plant height (Ht), which were later used to estimate the wood volume. The model identity test was applied to evaluate the feasibility of using a single equation to predict the wood volume of all the forest species involved in the study. The complete model had a better fit for the wood volume prediction of the six forest species involved in this study when compared to a single equation. As the tree species studied had different performances in terms of wood volume, species-specific equations should be used to study their growth curves.

Genetic improvement greatly contributed to the success of the forest industry in Brazil. While past selection efforts in Eucalyptus spp. have yielded satisfactory genetic gains, the response to selection in the last decade has fallen below expectations. Recent research suggests that inbreeding-based selection strategies, well established in crops such as rice and maize, could be adapted to enhance perennial species such as guava and can be expanded to forest species, such as eucalypt. In this context, 20 elite Eucalyptus spp. genotypes were self-pollinated, producing 30 progenies per family. A total of 600 individuals were planted in an experimental trial and evaluated for growth traits at 3 years of age. Both the progeny and the 20 parent genotypes were genotyped using SNP chips. This research aims to unravel the patterns of inbreeding depression in this S0:1 population of Eucalyptus spp., by examining autozygosity and genomic inbreeding patterns, estimating inbreeding depression (ID) and genetic parameters, and studying the impact of the unbalance between selfed and crossed individuals on the inbreeding depression estimator. The results revealed that dominance variance accounted for a notable portion of the phenotypic variation, demonstrating the significance of non-additive genetic effects in diameter at breast height (DBH). ID was evident, with reductions in DBH observed in most families as homozygosity increased. Simulations highlighted that unbalanced sample sizes of selfed and crossbred individuals could bias estimates of ID. By integrating genomic data and advanced quantitative methods, this study brings new information into the genetic consequences of self-pollination in eucalyptus, offering a foundation for managing inbreeding and enhancing genetic gains in perennial breeding. ### Competing Interest Statement The authors have declared no competing interest.

Micronuclei originate from DNA damage generated by clastogenic and/or by aneugenic effects. Depending on the pattern of damage, they may have distinct genomic origin and composition. Sequences of the centromere, telomere and rDNA have been identified in plant micronuclei. However, other DNA sequences may also be present in the micronuclei, as well as their DNA contents may be different. Here, we investigate the DNA content, genomic composition and origin of micronuclei induced in Zea mays by methyl methanesulfonate (MMS). DNA contents showed a wide range of distribution, suggesting their diverse genomic origins and illustrating how much of the nuclear genome can be lost due to mutagen effects. Micronuclei diversity was also evidenced by in situ probing with different DNA sequences (5S and 18S rDNAs, 180-bp knob and Grande LTR-retrotransposon) and by 6-diamidino-2 phenylindole (DAPI) fluorochrome. Perhaps these sequences are hotspots for MMS damage, especially the Grande LTR-retrotransposon, 5S and 18S rDNAs, which are rich in guanine. In addition, probe pools were constructed from individual genomic DNA of two microdissected micronuclei. These probe pools hybridized on all Z. mays chromosomes. However, the centromere, knob and secondary constriction were hybridized by only one probe pool, evidencing the distinct genomic composition of the micronuclei. We illustrate the micronuclei genomic diversity as they originated from several different chromosomes following the MMS treatment, and demonstrate the extent of the genotoxic damage to the genome. We provide some insights into micronuclei structure and diversity, and show that they can be further explored in mutagenesis research.

Optimizing seeding density is a fundamental strategy for maximizing wheat (Triticum aestivum L.) yield, yet establishing a universal recommendation is a persistent challenge in agronomy. The ideal plant population is highly dependent on complex interactions between genotype, management practices, and environmental conditions, often leading to contrasting results across studies. This variability makes generalized recommendations unreliable and highlights the critical need to understand cultivar-specific responses to guide precision management. The objective of this work was to test, using the model identity test, whether a single common regression model could adequately describe the cultivar responses or if distinct, genotype-specific models were required. A field experiment was conducted during summer and winter, in a randomized complete block design, and quadratic polynomial regression models were fitted for plant height, grain yield, hectoliter weight, and days to heading. The analysis of variance for the model identity test revealed significant differences (P < 0.05) between the complete (genotype-specific) and reduced (common) models for all traits in both seasons. This result led to the rejection of the null hypothesis of model equality, confirming that each cultivar exhibited a unique response pattern to the variation in seeding density. Similar genotype-specific and season-dependent responses were observed for all traits. The findings underscore strong genotypes × densities × environments interaction and demonstrate that a "one-size-fits-all" approach is inadequate for density recommendations. The identity test confirmed the need for distinct models, highlighting the importance of genotype-specific analysis in studies involving plant population density.

ABSTRACT This study aims to identify more relevant predictors traits, considering different prediction approaches in soybean under different shading levels in the field, using methodologies based on artificial intelligence and machine learning. The experiments were carried out under different shading levels in a greenhouse and in the field, using sixteen cultivars. We have evaluated grain yield, which was used as a response trait, and 22 other attributes as explanatory traits. Three levels of shading were used to restrict photosynthetically active radiation (RPAR): 0%, 25%, and 48%. At full sun level (0% RPAR), the traits that presented better predictive performances using a multilayer perceptron were specific leaf area, plant height and number of pods. In the three levels of shading, the plant height trait exhibited the best performance for the radial base function network. Plant height showed the best predictive efficiency for grain yield at 25% and 48% RPAR, for all machine learning methodologies. Computational intelligence and machine learning methodologies have proven to be efficient in predicting soybean grain yield, regardless of shading level.

The objective was to relate, based on quality characteristics, the potential of fifty-nine (59) sweet orange cultivars grafted onto the 'Sunki Tropical' tangerine rootstock and the 'Trifoliata' citrandarin hybrids 'San Diego', 'Riverside', and 'Indio', using 236 combinations with mixed models. The orchard was established in 2015. The following parameters were evaluated at stage III (fruit maturation): fruit weight (g), juice yield (JY, %), total soluble solids content (TSS, °Brix), titratable acidity (TA, g of citric acid per L-1), and the Ratio (TSS/TA). The mixed model methodology (REML/BLUP) was used to estimate/predict the fruit maturation parameters for different scion and rootstock combinations. The scion and rootstock combinations differed statistically for total soluble solids, titratable acidity, ratio, and juice yield. The rootstocks 'San Diego', 'Riverside', and 'Indio', and the 'Sunki Tropical' mandarin induced juice yields (55.05% to 60.05%), soluble solids (10.70°Brix to 12.38°Brix), and titratable acidity (0.50% to 0.66% citric acid) in sweet orange trees that were within quality standards. In the ranking, the combinations, in order, of 'Pera CNPMF D-9'/'Riverside', 'Pera Selection CNPMF C-32'/'Riverside', 'Pera Selection CNPMF D-3'/'San Diego', 'Pera Selection Olímpia'/'Sunki Tropical', 'Valência Selection CNPMF'/'Sunki Tropical', 'Diva'/'Sunki Tropical', 'Melrosa'/'Índio', 'Pera CNPMF D-6'/'Sunki Tropical', and 'Pera C-21'/'Riverside', scion and rootstock, respectively, stood out. These are combinations with greater potential for commercial exploitation because they meet quality standards.

This study focused on incorporating dimensionality reduction based on marker significance to better harness the potential of machine learning for genomic prediction in different trait-genomic structures. The aim was to show that outcomes achieved with reduced data would improve predictive accuracy ( ) and precision (root-mean-square error: RMSE) while reducing computational time. Distinct subsets of markers, in simulated data, were chosen by prioritizing importance via the Bagging technique. Predictive modelling was subsequently conducted using both Bagging and the diverse architectures of a Multilayer Perceptron (MLP) neural network. This study was carried out with six traits of an F2 simulated population (derived from contrasting homozygotes) with 1,000 individuals. Three traits had three different heritabilities (0.4, 0.6, and 0.8) and were controlled by a set of 40 quantitative trait loci (QTLs). Additionally, four QTLs with more pronounced heritability effects (set at unity) were introduced in three other traits while preserving the same genetic control structure as the earlier traits. In our investigation, as the number of markers increased, both techniques gradually increased training time; however, the time needed for computation notably extended beyond the threshold of 100 markers for Bagging. In comparison to the MLP model, the Bagging model generally obtained better accuracy (higher ) and precision (lower RMSE) values regardless of heritability and added QTLs. Most importantly, results highlight that for traits subject to robust genetic control of additional QTLs, MLP networks experienced a decline in prediction performance from a few markers (~10). In contrast, Bagging kept constant or subtly improved predication performance. Finally, the dimensionality reduction procedure effectively improves genomic prediction, and Bagging captures complex genetic control structures for prediction better than MLP networks.

In Brazil, disease outbreaks in plant cultivars are common in tropical zones. For example, the fungus Fusarium verticillioides produces mycotoxins called fumonisins (FUMO) which are harmful to human and animal health. Besides the genetic component, the expression of this polygenic trait is regulated by interactions between genes and environmental factors (G × E). Genomic selection (GS) emerges as a promising approach to address the influence of multiple loci on resistance. We examined different manners to conduct the prediction of FUMO contamination using genomic and pedigree data, and combinations of these two via the single step model (B-matrix) which also offers the possibility of increasing training set sizes. This is the first study to apply the B-matrix approach for predicting FUMO in tropical maize breeding programs. Our research introduced a cross-validation approach to optimize the hyper-parameter w, which represents the fraction of total additive variance captured by the markers. We demonstrated the importance of selecting optimal w by environment in unbalanced datasets. A total of 13 predictive models considering General Combining Ability (GCA) and Specific Combining Ability (SCA) effects, resulted from five linear predictors and three different covariance structures including the single-step approach. Two cross-validation scenarios were considered to evaluate the model's proficiency: CV1 simulated the prediction of completely untested hybrids, where the individuals in the validation set had no phenotypic records in the training set; and CV2 simulated the prediction of partially tested hybrids, where individuals had been evaluated in some environments but not in the target environment. Results showed that using the B-matrix in the five tested linear models increased the predictive ability compared to pedigree or genomic information. Under CV1, increasing training set sizes exhibit superior predictive accuracy. On the other hand, under CV2 the advantages of increasing the training set size are unclear and the improvements are due to better covariance structures. These insights can be applied to plant breeding programs where the GCA, SCA, and G × E interactions are of interest and pedigree information is accessible, but constraints related to genotyping costs for the entire population exist.

Species within the genus Corymbia are regarded as potential alternatives to Eucalyptus. In addition to having superior wood quality, Corymbia spp. are tolerant to most pests, diseases, and abiotic stresses that affecting Eucalyptus plantations, including physiological disorders, water deficit, and wind damage. However, environmental stresses stimulate kino production, which decreases the quality of pulp and sawn wood. This study aimed to develop a method for evaluating kinoand estimate genetic parameters in Corymbia. For this, 16 Corymbia (C. citriodora × C. torelliana) hybrid clones and 5 clones of Eucalyptus were used. Two evaluation methods (M1 and M2) were tested for kino evaluation; M1 consisted of drilling the bark with Pilodyn and M2 consisted of drilling the heartwood with Pilodyn. The following kino parameters were evaluated: exudation incidence, exudate length which flowed over the stem, and exudate weight. Genetic parameters were estimated by a mixed model method (REML/BLUP). The significance of random effects of the statistical model was tested by the likelihood ratio test. Significant clone effects were obtained for all kino parameters, except for exudate length as assessed by M2. Kino parameters determined by M1 exhibited higher heritability and accuracy. Therefore, M1 should be preferred for kino evaluation in Corymbia.

Open-pollinated onion cultivars predominate in the southern region of Brazil, due to their higher adaptability to local climatic conditions, unlike commercial hybrids, which have shown a lower adaptability. Cytoplasmic male sterility (CMS) systems are employed to develop hybrid onion cultivars. Two molecular markers, 5'cob and orfA501, were used to differentiate the cytoplasm type, and the AcSKP1 marker to identify the nuclear male fertility-restoring locus (Ms). A total of 1,126 plants from the most common onion cultivars grown in southern Brazil, including Bola Precoce (R), (R) , Crioula (R), (R) , Valessul (R), (R) , Mega (R), (R) , Joia (R) (R) and Princesa do Sul (R), (R) , were analyzed using all the three markers. An extremely rare occurrence of the S cytoplasm was observed among the cultivars, being detected in only 1.8 % of the samples, while the T cytoplasm was the most prevalent, accounting for 56.3 % of the samples. Among the 1,126 plants analyzed, only three exhibited the S cytoplasm and were recessive for the Ms-locus (Smsms). Additionally, 49 plants with the N-cytoplasm (as per the Engelke's classification) and recessive for the nuclear genotype (Nmsms) were identified, 45 of which were pollen producers. The male-fertility restoration occurred in 22.2 % of the crosses between Tmsms male-sterile plants and male-fertile N cytoplasmic-msms plants (as per the Engelke's classification).

In soybean breeding programs, a great deal of time is devoted to the use of methods that perform selection of individual plants during the initial generations. Our hypothesis is that BLUPIS (simulated individual BLUP) can be efficient when applied in the initial stages of soybean breeding programs. This study aimed to explore the potential of BLUPIS in the early generations of a soybean breeding program, as well as to assess the viability of the strategy of dividing the useful area of experimental plots for estimating genotypic effects and plant selection. The experiment involved 84 segregating populations and 15 soybean parents in the F2 and F3 generations. Yield data was collected from the 2019/2020 and 2020/2021 cropping seasons. In the F2 generation, different data exploration methods were applied to determine the most suitable adaptation to be used in the F3 generation. The individual BLUP (BLUPI) was compared with BLUPIS using information from different replications and/or equal to the information used in BLUPI. The selection conducted by BLUPIS and BLUPI showed high concordance regarding the selected plants. In the F3 generation, segregating populations were selected based on positive genotypic effects, and individual plants within these populations were further selected according to the number of plants determined by BLUPIS. The division of the plot area was an efficient strategy for selecting segregating populations and individual plants within superior populations in the F3 generation, resulting in genetic gains of approximately 1.56 g per plant. When combined with the strategy of advancing generations in the off-season, the BLUPIS approach reduces the time required to achieve a high level of homozygosity. Therefore, BLUPIS proved to be a powerful statistical tool for early selection based on grain yield in soybeans.

Selecting parents and crosses is a critical step for a successful breeding program. The ability to design crosses with high means that will maintain genetic variation in the population is the goal for long-term applications. Herein, we describe a new computational package for mate allocation in a breeding program. SimpleMating is a flexible and open-source R package originally designed to predict and optimize breeding crosses in crops with different reproductive systems and breeding designs. Divided into modules, SimpleMating first estimates the cross performance (criterion), such as mid-parental value, cross total genetic value, and/or usefulness of a set of crosses. The second module implements an optimization algorithm to maximize a target criterion while minimizing next-generation inbreeding. The software is flexible, enabling users to specify the desired number of crosses, set maximum and minimum crosses per parent, and define the maximum allowable parent relationship for creating crosses. As an outcome, SimpleMating generates a mating plan from the target parental population using single or multi-trait criteria. For example, we implemented and tested SimpleMating in a simulated maize breeding program obtained through stochastic simulations. The crosses designed via SimpleMating showed a large genetic mean over time (up to 22% more genetic gain than conventional genomic selection programs, with lesser loss of genetic diversity over time), supporting the use of this tool, as well as the use of data-driven decisions in breeding programs.

Understanding the genotype-by-environment interaction (GEI) and considering it in the selection process is a sine qua non condition for the expansion of Brazilian eucalyptus silviculture. This study's objective is to select high-performance and stable eucalyptus clones based on a novel selection index that considers the Factor Analytic Selection Tools (FAST) and the clone's reliability. The investigation explores the nuances interplay of GEI and extends its insights by scrutinizing the relationship between latent factors and real environmental features. The analysis, conducted across seven trials in five Brazilian states involving 78 clones, employs FAST. The clonal selection was performed using an extended FAST index weighted by the clone's reliability. Further insights about GEI emerge from the integration of factor loadings with 25 environmental features through a principal component analysis. Ten clones, distinguished by high performance, stability, and reliability, have been selected across the target population of environments. The environmental features most closely associated with factor loadings, encompassing air temperature, radiation, and soil characteristics, emerge as pivotal drivers of GEI within this dataset. This study contributes insights to eucalyptus breeders, equipping them to enhance decision-making by harnessing a holistic understanding-from the genotypes under evaluation to the diverse environments anticipated in commercial plantations.

Genomic selection and doubled haploids hold significant potential to enhance genetic gains and shorten breeding cycles across various crops. Here, we utilized stochastic simulations to investigate the best strategies for optimize a sweet corn breeding program. We assessed the effects of incorporating varying proportions of old and new parents into the crossing block (3:1, 1:1, 1:3, and 0:1 ratio, representing different degrees of parental substitution), as well as the implementation of genomic selection in two distinct pipelines: one calibrated using the phenotypes of testcross parents (GSTC scenario) and another using F1 individuals (GSF1). Additionally, we examined scenarios with doubled haploids, both with (DH) and without (DHGS) genomic selection. Across 20 years of simulated breeding, we evaluated scenarios considering traits with varying heritabilities, the presence or absence of genotype-by-environment effects, and two program sizes (50 vs 200 crosses per generation). We also assessed parameters such as parental genetic mean, average genetic variance, hybrid mean, and implementation costs for each scenario. Results indicated that within a conventional selection program, a 1:3 parental substitution ratio (replacing 75% of parents each generation with new lines) yielded the highest performance. Furthermore, the GSTC model outperformed the GSF1 model in enhancing genetic gain. The DHGS model emerged as the most effective, reducing cycle time from 5 to 4 years and enhancing hybrid gains despite increased costs. In conclusion, our findings strongly advocate for the integration of genomic selection and doubled haploids into sweet corn breeding programs, offering accelerated genetic gains and efficiency improvements.

The assessment of plant performance and the accuracy of genetic selection can be significantly affected by genetic competition among individuals. In addition to genetic causes, competition is influenced by external factors such as environment and age. This research uses multi-location multi-age Eucalyptus dunii trials to answer four questions: i) Are there major changes when competition effects (both genetic and residual) are estimated in single-age and multi-age models? ii) What are the implications of considering competition effects on the early selection of eucalypt clones? iii) What are the impacts of considering the reliability of direct (DGE) and indirect genotypic effects (IGE) as a weight in a selection index?, and iv) Which clones hold the potential to form highly productive clonal plantations in Southern Brazil when deployed together as clonal composites? The dataset contained three trials established in different locations in Southern Brazil, where growth traits were measured at 3.5 and 7 years. We fitted single-age and multi-age spatial competition models for each trial. The multi-age spatial competition model was a valuable tool for selecting superior eucalypt clones. Neglecting the IGEs led to changes in clone ranking, reducing the effectiveness of early selection. Additionally, a selection index where DGEs and IGEs are weighted by their reliability was proposed. Therefore, our study contributes to understanding the magnitude of IGE's impact on the daily practice of eucalypt breeding, such as early selection and multi-age analyses, and proposes selection strategies that consider the quantity and quality of information provided by the models for each individual.