GenomicLand: A user-friendly tool for genomic prediction and association studies

Abstract

GenomicLand is a free, open-source platform for genomic prediction and genome-wide association studies. Developed in R software using Shiny, it supports diploid and polyploid datasets and integrates statistical learning, Bayesian, machine learning, and neural network methods. GenomicLand provides an accessible environment for researchers, breeders, and students.

Keywords:
Mixed models; statistical learning; machine learning; Bayesian inference; plant breeding

INTRODUCTION

GenomicLand is a free software platform developed for genomic prediction and genome-wide association studies (GWAS). The first version was released in 2019 (Azevedo et al. 2019), combining R for analytical methods and Python for the graphical user interface. The current version has been fully migrated to R using the Shiny framework, improving accessibility, integration, and usability. GenomicLand provides a unified graphical interface compatible with Windows and Linux operating systems and supports phenotypic and genomic data from diploid and polyploid species.

Genome-wide selection (GWS) and GWAS are key tools in modern breeding programs. These approaches improve the accuracy of genetic value prediction (Meuwissen et al. 2001, Wellmann and Bennewitz 2012), accelerate the identification of superior genotypes (Vandenplas et al. 2018), increase genetic gain, and reduce generation intervals (Resende et al 2012). Additionally, they support parental selection by identifying individuals with desirable genetic profiles for future crosses. GWAS also contributes to understanding the genetic architecture of complex traits by identifying genomic regions associated with phenotypic variation (Uffelmann et al. 2021). However, implementing these methodologies often requires programming expertise and the integration of multiple tools, which may limit their accessibility to a broader user community.

The availability of free and user-friendly software is essential for expanding access to advanced genomic analyses. GenomicLand addresses this gap by integrating statistical learning, mixed models, Bayesian inference, machine learning, and artificial intelligence methods into a single analytical environment, enabling genomic analyses without requiring advanced programming skills.

The current version expands its capabilities by supporting additive and non-additive models, diploid and polyploid species, genomic prediction, GWAS, genotype-by-environment interaction analyses, and multivariate approaches for multi-trait and multi-environment data.

SOFTWARE ARCHITECTURE

GenomicLand was developed entirely in the R programming language using the Shiny framework. The software architecture adopts a modular architecture that integrates data processing, statistical modeling, and visualization within a unified analytical environment (Figure 1A).

Figure 1
GenomicLand graphical interface, analytical workflow, and data preprocessing module. (A) Main graphical user interface showing the analytical modules available in the software. (B) Analytical workflow from data input and preprocessing to genomic analyses and output generation. (C) Data preprocessing module, including quality control, imputation, allelic diversity, relationship matrix construction, linkage disequilibrium analysis, and SNP density visualization.

The system architecture comprises three main layers: i) a user interface layer, which allows users to upload datasets, configure model parameters, and visualize analytical results through a graphical interface; ii) a computational layer, responsible for executing statistical and machine learning analyses using established R packages and custom routines; and iii) a visualization layer, which generates graphical outputs such as heatmaps, GWAS Manhattan plots, linkage disequilibrium (LD) decay curves, and model diagnostic plots.

GenomicLand is available for Windows and Linux operating systems; it is freely accessible at https://www.licae.ufv.br/genomicland/. The software supports genotype and phenotype datasets in standard tabular formats (e.g., CSV, TXT, and XLSX files) and provides example datasets to facilitate testing and reproducibility.

This architecture facilitates the seamless integration of multiple analytical approaches, thereby reducing the need for external tools and simplifying complex genomic analyses for users with varying levels of programming experience. This integrated design distinguishes GenomicLand from single-purpose tools by providing an integrated environment for end-to-end genomic analyses.

INPUT DATA STRUCTURE

GenomicLand requires genotype and phenotype data in standardized tabular formats to ensure compatibility across all analytical modules.

Genotype data

The genotype file is organized with genetic markers (single nucleotide polymorphisms, SNPs) in rows and individuals in columns. The first columns contain marker information, including the SNP identifier, chromosome, and genomic position, expressed either as a physical (base pairs) or genetic (centiMorgans) position. The remaining columns correspond to individual genotypes.

Phenotype data

The phenotype file is organized with individuals in rows. The first column contains individual identifiers matching those in the genotype file, whereas subsequent columns contain phenotypic traits and optional fixed and random effects.

This standardized structure ensures compatibility across analytical modules and facilitates genomic analyses within the software.

AVAILABLE PROCEDURES

GenomicLand comprises a comprehensive set of analytical modules for data processing, GWS, and GWAS (Figure 1A). These modules are integrated within a unified analytical environment and can be accessed through the main navigation menu, enabling users to perform exploratory analyses, model fitting, prediction, and statistical inference within a structured workflow (Figure 1B).

Data preprocessing

The data preprocessing module provides tools for genotype data preparation, quality assessment, and exploratory analyses prior to genomic analyses (Figure 1C), including:

1. Allelic dosage: Converts nucleotide genotypes into numerical allele dosage values ranging from 0 to the ploidy level. The most frequent nucleotide is designated as the reference (major) allele, whereas the less frequent nucleotide is considered the alternative (minor) allele. This encoding follows standard practices in genomic prediction and facilitates downstream statistical modeling.

2. Quality control: Enables the removal of markers with low minor allele frequency (MAF) and low call rate and individuals with excessive proportions of missing data.

3. Imputation: Missing allelic dosage values can be imputed using the marker incidence mean, mode, or a random imputation approach.

4. SNP density plot: Visualizes SNP distribution across the genome and highlights regions with varying marker density. Graphical outputs are generated using CMplot (LiLin-Yin 2024).

5. Allelic diversity: Estimates allele frequencies, polymorphic information content (PIC), observed heterozygosity, and expected heterozygosity.

6. Relationship matrix (G): Computes genomic relationship matrices for diploid and polyploid species. The module also generates heatmaps and principal component analysis (PCA) plots to support the evaluation of population structure and genomic relationships among individuals.

7. Linkage disequilibrium (LD): Estimates LD decay using marker data and genetic or physical distances. LD is quantified by the squared correlation coefficient (r²) between marker pairs and summarized through LD decay curves.

Interface and input configuration

As illustrated in Figure 1C, the data preprocessing module provides an integrated interface for defining input parameters and executing preprocessing procedures. Required inputs include the genotype file, missing-data coding scheme, ploidy level, output directory, and filtering thresholds.

Three standard actions are available throughout the software: Process Data, Load Example, and Clean Input. Users can execute analyses using their own datasets or explore the software functionality through internal example datasets. Each module includes specific input parameters and generates tailored outputs according to the selected procedure. These procedures contribute to data quality and reliability, supporting accurate and robust downstream genomic analyses.

Phenotype analysis

Phenotype correction: Phenotypic data can be adjusted for fixed and random effects, including categorical factors, numeric covariates, and population structure, via principal components (Azevedo et al. 2017). Depending on the analytical design, adjustments can be performed using linear regression or mixed models to obtain genotype-level best linear unbiased estimators (BLUEs; Holland and Piepho 2024) for downstream genomic analyses (Figure 2). All analyses are implemented using the R package sommer (Covarrubias-Pazaran 2016).

Figure 2
Overview of the analytical procedures implemented in GenomicLand. The figure summarizes the main analytical modules available in GenomicLand for genomic prediction and association studies. The software is organized into five major components: (1) Data preprocessing, including allelic dosage conversion, quality control, missing data imputation, SNP density visualization, allelic diversity analysis, genomic relationship matrix construction, and linkage disequilibrium (LD) analysis; (2) Phenotype analysis, including phenotype correction through fixed and random effects, covariates, and population structure adjustment; (3) Genomic prediction, encompassing regularization and mixed models, Bayesian methods, dimensionality reduction approaches, machine learning algorithms, neural networks, non-additive models, genotype-by-environment (G×E) models, and multi-trait analyses; (4) Genome-wide association studies (GWAS), including single-marker, regional heritability mapping, and Bayesian GWAS approaches; and (5) Non-additive GWAS, comprising additive, dominance, and epistatic association models.

Genomic prediction methods

The genomic prediction modules provide outputs including genomic estimated values, heritability estimates, marker effects, goodness-of-fit statistics, and predictive ability and bias assessed through K-fold cross-validation.

Regularization and mixed models

This module implements genomic prediction methods based on regression coefficient shrinkage and linear mixed models (Figure 2). Available methods include least absolute shrinkage and selection operator (LASSO), pedigree-based best linear unbiased prediction (A-BLUP), genomic BLUP (G-BLUP), single-step BLUP (H-BLUP), and heterogeneous G-BLUP.

1. Least absolute shrinkage and selection operator (LASSO): LASSO performs simultaneous variable selection and regularization through a shrinkage parameter (λ), optimized by K-fold cross-validation, producing sparse models by shrinking some coefficients to zero (Tibshirani 1996).

2. Mixed model methods: This module implements A-BLUP, G-BLUP, and H-BLUP based on pedigree (A), genomic (G), or combined (H) relationship matrices (Henderson 1975, VanRaden 2008, Legarra et al. 2014). These methods fit linear mixed models that accommodate fixed and random effects, including categorical factors and continuous covariates. Variance components are estimated by restricted maximum likelihood, and breeding values are obtained through BLUP. The module supports heterogeneous G-BLUP, which incorporates marker-specific weights when constructing the G matrix (Su et al. 2014). All models are fit using the R package sommer (Covarrubias-Pazaran 2016), and relationship matrices are computed using AGHmatrix (Amadeu et al. 2023).

Bayesian methods

This module implements marker-based Bayesian genomic prediction models that differ in their prior assumptions regarding marker effects and kernel-based approaches for estimating genomic values (Pérez and de los Campos 2014).

1. Bayesian ridge regression (BRR): Assumes homogeneous shrinkage, such that all marker effects are shrunk toward zero equally.

2. BayesA: Assumes heterogeneous shrinkage, allowing each marker to have its own variance and effect-specific shrinkage.

3. BayesB: Assumes that a user-defined proportion (π) of markers has no effect, whereas the remaining markers are modeled using heterogeneous shrinkage.

4. BayesCπ: Combines a point mass at zero with homogeneous shrinkage, while estimating the proportion of markers with non-zero effects from the data.

5. BayesDπ: Extends BayesCπ by assuming heterogeneous shrinkage for markers with non-zero effects.

6. Bayesian least absolute shrinkage and selection operator (BLASSO): Applies enhanced heterogeneous shrinkage to marker effects, favoring sparse solutions.

7. Reproducing kernel Hilbert space (RKHS): A non-parametric Bayesian approach that uses kernel matrices to model complex and potentially non-linear relationships between markers and phenotypes, allowing the capture of epistatic and other non-additive effects.

All Bayesian models are fit using the R package BGLR (Pérez and de los Campos 2014).

Dimensionality reduction methods

These methods transform high-dimensional predictors (e.g., marker incidence matrices) into a reduced set of latent components, improving computational efficiency and reducing collinearity (Azevedo et al. 2013). The optimal number of components is determined by K-fold cross-validation.

1. Principal component regression (PCR): Constructs uncorrelated components that maximize the variance explained by the predictor variables.

2. Partial least squares regression (PLSR): Constructs components that maximize the covariance between predictors and the response variable.

3. Independent component regression (ICR): Constructs statistically independent components to extract informative latent signals for prediction.

PCR and PLSR models are fit using the R package pls (Liland et al. 2024). ICR models are fit using functions from the R package caret (Kuhn 2008), combined with a custom ICR implementation based on Azevedo et al. (2013).

Machine learning methods

This module implements machine learning approaches for genomic prediction that can capture complex and potentially non-linear relationships between genetic markers and phenotypes (González-Camacho et al. 2018, Sousa et al. 2020, Oliveira et al. 2021, Oliveira et al. 2024).

1. Bootstrap aggregating (Bagging): Ensemble method that combines predictions from multiple regression trees fitted to bootstrap samples of the training data. Implemented using the R package ipred (Peters and Hothorn 2024).

2. Random Forest: Extension of Bagging that incorporates random subsets of predictors at each tree split to improve predictive performance. Implemented using the R package randomForest (Liaw and Wiener 2002).

4. Boosting: Sequential ensemble learning method in which trees are fitted iteratively to reduce prediction errors from previous iterations. Implemented using the R package gbm (Ridgeway and Developers 2026).

5. Multivariate adaptive regression splines (MARS): Non-parametric regression approach that models non-linear effects and interactions through piecewise linear splines. Implemented using the R package earth (Milborrow et al. 2024).

6. Quantile regression: Estimates predictor effects at user-defined quantiles of the response distribution, providing robustness to outliers and heterogeneous variances. Implemented using the R package quantreg (Koenker 2025).

Neural network methods

This module implements neural network models for genomic prediction that are particularly suitable for capturing complex non-linear and interaction effects among genetic markers (Cruz and Nascimento 2018, Costa et al. 2022).

Multilayer perceptron (MLP): (MLP): Feedforward neural network with one or more hidden layers for modeling complex non-linear relationships between predictors and response variables. Users can define the number of hidden layers, neurons, and activation functions. Models are implemented using neuralnet (Fritsch et al. 2019), with variable importance assessed using NeuralNetTools (Beck 2018) and model evaluation performed through K-fold cross-validation using caret (Kuhn 2008).

Non-additive models

This module extends the G-BLUP framework to include dominance and epistatic interactions, including additive × additive, additive × dominance, and dominance × dominance effects (Wellmann and Bennewitz 2011, Vitezica et al. 2013). By accounting for both additive and non-additive sources of genetic variation, these models provide a more comprehensive representation of the genetic architecture of complex traits. The significance of non-additive effects is evaluated using likelihood ratio tests. A Bayesian version of G-BLUP is also available with the same interaction structures, enabling model comparison based on the deviance information criterion (Wang et al. 2004, Wellmann and Bennewitz 2012, Azevedo et al. 2015). The module additionally includes marker-based Bayesian models for the joint modeling of additive and dominance effects.

Genotype-by-environment (G×E) models

This module evaluates genotype performance across multiple environments by explicitly modeling genotype × environment interactions, enabling the identification of genotypes with stable performance or specific adaptability (Resende et al. 2014).

1. BG-BLUP GE: A Bayesian extension of the G-BLUP framework that incorporates environmental effects, genomic relationships, and G×E interactions within a linear mixed-model framework. The significance of G×E effects can be evaluated using likelihood ratio tests.

2. Multi-BG-BLUP GE: A multivariate Bayesian model for multi-environment data that jointly analyzes observations across environments, enabling the estimation of genetic correlations and potentially improving predictive performance by borrowing information across environments.

Multi-trait models

This module implements multivariate Bayesian linear mixed models for the joint analysis of multiple traits. The models accommodate fixed and random effects while estimating genetic correlations among traits. By leveraging information shared across correlated traits, these approaches can improve prediction accuracy and the estimation of genetic parameters.

Genome-wide association methods

This module implements GWAS methods based primarily on additive genetic effects, supporting analyses at both the individual-marker and genomic-region levels. These approaches enable the identification of genomic regions associated with phenotypic variation while accounting for population structure and genetic relatedness.

1. Single marker method (SMM): Individual SNPs are tested sequentially while controlling population structure through PCA and background polygenic effects using the genomic relationship matrix (G). The method adopts a leave-one-chromosome-out (LOCO) strategy, in which chromosome-specific genomic relationship matrices are constructed excluding the chromosome under evaluation, reducing confounding between marker and polygenic effects. Analyses are implemented using the R package GWASpoly (Rosyara et al. 2016).

2. Regional heritability mapping (RHM): Estimates the heritability of predefined genomic regions, capturing the cumulative additive effects of markers within each segment (Suela et al. 2022). Population structure is accounted for through polygenic effects modeled using the genomic relationship matrix (G), and region size is defined by the user.

3. Alphabetic Bayesian methods: Bayesian GWAS analyses are performed using genomic prediction models under different prior assumptions. Genomic regions are identified through the window posterior probability of association (WPPA), which quantifies the proportion of genetic variance explained by markers within a genomic window (Fernando et al. 2017, Lima et al. 2022). Significance thresholds are user-defined.

Non-additive GWAS

This module extends the GWAS approaches described above by incorporating dominance effects, enabling the partitioning of genetic variance into additive and dominance components (Rosyara et al. 2016, Azevedo et al. 2022). By explicitly modeling non-additive genetic effects, these approaches may improve the detection of marker-trait associations when dominance contributes substantially to phenotypic variation.

Application example

To demonstrate the functionality of GenomicLand, a publicly available dataset of Oryza sativa L. (Asian rice) was analyzed. The dataset comprised 413 individuals and 36,901 SNP markers generated through the OryzaSNP and OMAP projects (Ammiraju et al. 2006, Zhao et al. 2011).

Plant height was selected to illustrate the analytical workflow. Representative outputs included marker quality and genomic structure summaries (Figure 3A-D), genomic prediction results (Figure 3E-H), and genome-wide association analyses (Figure 3I-L). Marker quality assessment revealed adequate allele frequency distributions, linkage disequilibrium patterns, and genome-wide marker coverage. Genomic prediction analyses enabled the comparison of alternative methods in terms of predictive ability, heritability, prediction error, and computational time. GWAS analyses identified putative genomic regions associated with plant height and provided graphical outputs, including Manhattan plots, QQ plots, and genotype-specific phenotype distributions.

Figure 3
Representative outputs generated by GenomicLand using the Oryza sativa dataset. The figure summarizes the main analytical workflow available in GenomicLand, from data preprocessing to genomic prediction and genome-wide association analyses. (A) Distribution of minor allele frequency (MAF) across SNP markers. (B) Distribution of polymorphic information content (PIC). (C) Linkage disequilibrium (LD) decay represented by the decline of the squared correlation coefficient (r²) as a function of physical distance (Mb). (D) SNP density across chromosomes. (E) Predictive ability, (F) heritability (h²), (G) mean squared error (MSE), and (H) computational time obtained using different genomic prediction methods. (I) Manhattan plot of GWAS results for plant height under the additive model. (J) Chromosome-specific quantile-quantile (QQ) plots comparing observed and expected −log₁₀(p) values. (K-L) Phenotypic distributions according to genotype classes for representative SNPs associated with plant height.

GenomicLand was developed primarily for educational and research purposes. The analyses presented here were performed on a workstation equipped with an Intel Core i7-13700 processor and 64 GB of RAM. Computational requirements depend on dataset size and analytical complexity, particularly for Bayesian and machine learning methods. Future developments will focus on improving computational efficiency and expanding analytical capabilities for genomic datasets.

ACKNOWLEDGMENTS

The authors acknowledge the financial support provided by the Foundation for Research Support of the State of Minas Gerais (FAPEMIG) through projects APQ-02111-18 and APQ-01545-24 and by the Brazilian National Council for Scientific and Technological Development (CNPq) through projects 309856/2023-0, 310755/2023-9, and 408833/2023-8. The authors also acknowledge, in memoriam, Dr. Fabyano Fonseca e Silva for his important contributions during the early development of this project, his role in shaping the software, and for naming it GenomicLand.

REFERENCES

• Amadeu RR, Garcia AAF, Munoz PR, Ferrão LFV2023 AGHmatrix: genetic relationship matrices in R. Bioinformatics 39: btad445.
• Ammiraju JS, Luo M, Goicoechea JL, Wang W, Kudrna D, Mueller C, Talag J, Kim H, Sisneros NB, Blackmon B, Fang E, Tomkins JB, Brar D, MacKill D, McCouch S, Kurata N, Lambert G, Galbraith DW, Arumuganathan K, Rao K, Walling JG, Gill N, Yu Y, SanMiguel P, Soderlund C, Jackson S, Wing RA2006 The Oryza bacterial artificial chromosome library resource: construction and analysis of 12 deep-coverage large-insert BAC libraries that represent the 10 genome types of the genus OryzaGenome Research 16:140-147
• Azevedo CF, Lima LP, Nascimento M, Nascimento ACC2022 Bayesian methods for genomic association of chromosomic regions considering the additive-dominance modelCrop Breeding and Applied Biotechnology 22:e418422310
• Azevedo CF, Nascimento M, Fontes VCF, Silva FF, Resende MDV, Cruz CD2019 GenomicLand: Software for genome-wide association studies and genomic prediction. Acta ScientiarumAgronomy 41:e45361
• Azevedo CF, Resende MDV, Silva FF, Lopes OS, Guimarães SEF2013 Regressão via componentes independentes aplicada à seleção genômica para características de carcaça em suínosPesquisa Agropecuária Brasileira 48:619-626
• Azevedo CF, Resende MDV, Silva FF, Nascimento M, Viana JMS, Valente MSF2017 Population structure correction for genomic selection through eigenvector covariatesCrop Breeding and Applied Biotechnology 17:350-358
• Azevedo CF, Resende MDV, Silva FF, Viana JMS, Valente MSF, Resende Junior MFR, Muñoz P2015 Ridge, Lasso and Bayesian additive-dominance genomic modelsBMC Genetics 16:1-13
• Beck MW2018 NeuralNetTools: Visualization and analysis tools for neural networksJournal of Statistical Software 85:1-20
• Costa WG, Celeri MO, Barbosa IP, Silva GN, Azevedo CF, Borem A, Nascimento M, Cruz CD2022 Genomic prediction through machine learning and neural networks for traits with epistasisComputational and Structural Biotechnology Journal 20:5490-5499
• Covarrubias-Pazaran G2016 Genome-assisted prediction of quantitative traits using the R package sommerPLoS ONE 11:e0156744
• Cruz CD, Nascimento M2018 Inteligência computacional aplicada ao melhoramento genético. Editora UFV, Viçosa, 414p.
• Fernando R, Toosi A, Wolc A, Garrick D, Dekkers J2017 Application of whole-genome prediction methods for genome-wide association studies: a Bayesian approachJournal of Agricultural, Biological and Environmental Statistics 22:172-193
• Fritsch S, Guenther F, Wright M2019 neuralnet: Training of neural networks. R package version 1.44.2.
• González-Camacho JM, Ornella L, Pérez-Rodríguez P, Gianola D, Dreisigacker S, Crossa J2018 Applications of machine learning methods to genomic selection in breeding wheat for rust resistanceThe Plant Genome 11:1-15
• Henderson CR1975 Best linear estimation and prediction under a selection modelBiometrics 31:423-447
• Holland JB, Piepho HP2024 Don't BLUP twice. G3 Genes | Genomes | Genetics 14: jkae250.
• Koenker R2025 quantreg: Quantile regression. R package version 6.1.
• Kuhn M2008 Building predictive models in R using the caret packageJournal of Statistical Software 28:1-26
• Legarra A, Christensen OF, Aguilar I, Misztal I2014 Single Step, a general approach for genomic selectionLivestock Science 166:54-65
• Liaw A, Wiener M2002 Classification and regression by randomForestR News 2:18-22
• Liland K, Mevik B, Wehrens R2024 pls: Partial least squares and principal component regression. R package version 2.8-5.
• LiLin-Yin2024 CMplot: Circle manhattan plot. R package version 4.5.1.
• Lima LP, Azevedo CF, Resende MDV, Nascimento M, Silva FF2022 Evaluation of Bayesian methods of genomic association via chromosomic regions using simulated dataScientia Agricola 79:3
• Meuwissen TH, Hayes BJ, Goddard ME2001 Prediction of total genetic value using genome-wide dense marker mapsGenetics 157:1819-1829
• Milborrow S, Hastie T, Tibshirani R2024 earth: Multivariate adaptive regression splines. R package version 5.3.4.
• Oliveira CM, Costa WG, Nascimento ACC, Azevedo CF, Cruz CD, Sagae VS, Nascimento M2024 Multivariate adaptive regression splines enhance genomic prediction of non-additive traitsAgronomy 14:2234
• Oliveira GF, Nascimento ACC, Nascimento M, Sant'Anna IC, Romero JV, Azevedo CF, Bhering LL, Caixeta ET2021 Quantile regression in genomic selection for oligogenic traits in autogamous plants: A simulation studyPLoS ONE 16:e0243666
• Pérez P, de los Campos G2014 Genome-wide regression and prediction with the BGLR statistical packageGenetics 198:483-95
• Peters A, Hothorn T2024 ipred: Improved predictors_. R package version 0.9-15.
• Resende MDV, Resende Junior MFR, Sansaloni CP, Petroli CD, Missiaggia AA, Aguiar AM, Abad JM, Takahashi EK, Rosado AM, Faria AD, Pappas Junior GJ, Kilian A, Grattapaglia D2012 Genomic selection for growth and wood quality in Eucalyptus: capturing the missing heritability and accelerating breeding for complex traits in forest treesNew Phytologist 194:116-128
• Resende MDV, Silva FF, Azevedo CF2014 Estatística matemática, biométrica e computacional. Editora Suprema, Visconde do Rio Branco, 881p.
• Ridgeway G, Developers G2026 gbm: Generalized boosted regression models. R package version 2.2.3.
• Rosyara UR, Jong WS, Douches DS, Endelman JB2016 Software for genome-wide association studies in autopolyploids and its application to potatoPlant Genome 9:1-10
• Sousa IC, Nascimento M, Silva GN, Nascimento ACC, Cruz CD, Silva FF, Almeida DP, Pestana KN, Azevedo CF, Zambolim L, Caixeta ET2020 Genomic prediction of leaf rust resistance to Arabica coffee using machine learning algorithmsScientia Agricola 78:e20200021
• Su G, Christensen OF, Janss L, Lund MS2014 Comparison of genomic predictions using genomic relationship matrices built with different weighting factors to account for locus-specific variancesJournal of Dairy Science 97:6547-6559
• Suela MM, Azevedo CF, Nascimento M, Nascimento ACC, Resende MDV2022 Regional heritability mapping and genome-wide association identify loci for rice traitsCrop Science 62:839-858
• Tibshirani R1996 Regression shrinkage and selection via the LassoJournal of the Royal Statistical Society, Series B (Methodological) 58:267-288
• Uffelmann E, Huang QQ, Munung NS, Vries J, Okada Y, Martin AR, Martin HC, Lappalainen T, Posthuma D2021 Genome-wide association studiesNature Reviews Methods Primers 1:59
• Vandenplas J, Calus MPL, Gorjanc G2018 Genomic Prediction using individual-level data and summary statistics from multiple populationsGenetics 210:53-69
• VanRaden PM2008 Efficient methods to compute genomic predictionsJournal of Dairy Science 91:4414-1423
• Vitezica ZG, Varona L, Legarra A2013 On the additive and dominant variance and covariance of individuals within the genomic selection scopeGenetics 195:1223-1230
• Wang J, van Ginkel M, Trethowan R, Ye G, DeLacy IH, Podlich DW, Cooper M2004 Simulating the effects of dominance and epistasis on selection response in the CIMMYT Wheat Breeding Program using QuCimCrop Science 44:2006-2018
• Wellmann R, Bennewitz J2011 The contribution of dominance to the understanding of quantitative genetic variationGenetics Research 93:139-154
• Wellmann R, Bennewitz J2012 Bayesian models with dominance effects for genomic evaluation of quantitative traitsGenetics Research 94:21-37
• Zhao K, Tung CW, Eizenga GC, Wright MH, Ali ML, Price AH, Norton JG, Islam AR, Reynolds A, Mezey J, Mcclung AM, Bustamante CD, Mcclung AM2011 Genome-wide association mapping reveals a rich genetic architecture of complex traits in Oryza sativaNature communications 2:467

Correspondence

Corresponding author: [e-mail]

SCIENTIFIC EDITOR:

Luiz Antônio dos Santos Dias

Publication Dates

• Publication in this collection
  25 Sept 2026
• Date of issue
  2026

History

• Received
  20 Jan 2026
• Accepted
  25 May 2026
• Published
  16 June 2026