Genome-enabled prediction through quantile random forest for complex traits / Predição genômica por meio do random forest quantílitico para características complexas

ABSTRACT: Quantile Random Forest (QRF) is a non-parametric methodology that combines the advantages of Random Forest (RF) and Quantile Regression (QR). Specifically, this approach can explore non-linear functions, determining the probability distribution of a response variable and extracting information from different quantiles instead of just predicting the mean. This evaluated the performance of the QRF in the genomic prediction for complex traits (epistasis and dominance). In addition, compare the accuracies obtained with those derived from the G-BLUP. The simulation created an F2 population with 1,000 individuals and genotyped for 4,010 SNP markers. Besides, twelve traits were simulated from a model considering additive and non-additive effects, QTL (Quantitative trait loci) numbers ranging from eight to 120, and heritability of 0.3, 0.5, or 0.8. For training and validation, the 5-fold cross-validation approach was used. For each fold, the accuracies of all the proposed models were calculated: QRF in five different quantiles and three G-BLUP models (additive effect, additive and epistatic effects, additive and dominant effects). Finally, the predictive performance of these methodologies was compared. In all scenarios, the QRF accuracies were equal to or greater than the methodologies evaluated and proved to be an alternative tool to predict genetic values in complex traits.

RESUMO: Quantile Random Forest (QRF) é uma metodologia não paramétrica, que combina as vantagens do Random Forest (RF) e da Regressão Quantílica (QR). Especificamente, essa abordagem pode explorar funções não lineares, determinando a distribuição de probabilidade de uma variável resposta e extraindo informações de diferentes quantis em vez de apenas prever a média. O objetivo deste trabalho foi avaliar o desempenho do QRF em predizer o valor genético genômico para características com arquitetura genética não aditiva (epistasia e dominância). Adicionalmente, as acurácias obtidas foram comparadas com aquelas advindas do G-BLUP. A simulação criou uma população F2 com 1.000 indivíduos genotipados para 4.010 marcadores SNP. Além disso, doze características foram simuladas a partir de um modelo considerando efeitos aditivos e não aditivos, com número de QTL (Quantitative trait loci) variando de oito a 120 e herdabilidade de 0,3, 0,5 ou 0,8. Para treinamento e validação foi usada a abordagem da validação cruzada 5-fold. Para cada um dos folds foram calculadas as acurácias de todos os modelos propostos: QRF em cinco quantis diferentes e três modelos do G-BLUP (com efeito aditivo, aditivo e epistático, aditivo e dominante). Por fim, o desempenho preditivo dessas metodologias foi comparado. Em todos os cenários, as acurácias do QRF foram iguais ou superiores às metodologias avaliadas e mostrou ser uma ferramenta alternativa para predizer valores genéticos em características complexas.

Overview of the use of genomic data in animal breeding / Panorama do uso de dados genômicos no melhoramento genético animal

ABSTRACT: The use of molecular information in breeding programs contributed to important advances in the improvement of traits of economic interest in livestock production. The advent of single nucleotide polymorphism (SNP) panels applied to genome-wide selection (GWS) and genome-wide association studies (GWAS), along with computational advances (e.g., use of powerful software and robust analyses) allowed a better understanding of the genetic architecture of farm animals and increased the selection efficiency. In this context, the statistic method single-step GBLUP has been frequently used to perform GWS, and more recently GWAS analyses, providing accurate predictions and QTL detection, respectively. Nevertheless, in developing countries, species such as sheep and goats, whose genomic data are more difficult to be obtained, the use of data simulation has been efficient in the study of the major factors involved in the selection process, such as size of training population, density of SNP chips, and genotyping strategies. The effects of these factors are directly associated with the prediction accuracy of genomic breeding values. In this review we showed important aspects of the use of genomics in the genetic improvement of production traits of animals, the main methods currently used for prediction and estimation of molecular marker effects, the importance of data simulation for validation of those methods, as well as the advantages, challenges and limitations of the use of GWS and GWAS in the current scenario of livestock production.

RESUMO: Em programas de melhoramento genético, o uso de informações moleculares garantiu importantes avanços para a melhoria de características de interesse econômico, no âmbito da produção animal. O advento da tecnologia de painéis de SNPs aplicados à seleção genômica ampla (GWS) e associação genômica ampla (GWAS), aliado ao avanço computacional, com o uso de softwares e análises robustas, permitiram melhor compreensão sobre a arquitetura genética dos animais de produção e, consequentemente, maior eficiência na seleção. Nesse contexto, o método estatístico single-step GBLUP tem sido utilizado, frequentemente, na execução da GWS e, mais recentemente, em GWAS, possibilitando predições acuradas e detecção de QTLs, respectivamente. No entanto, em países em desenvolvimento e, em espécies como os ovinos e caprinos, que existe maior dificuldade para a aquisição de dados genômicos, o uso da simulação de dados tem se mostrado eficiente para estudar os principais fatores envolvidos no processo de seleção, como o tamanho da população de treinamento, densidade de chipde SNPs e estratégias de genotipagem, cujos efeitos estão diretamente associados à acurácia da predição de valores genéticos genômicos. Nesta revisão, serão abordados pontos importantes sobre o uso da genômica no melhoramento genético de características produtivas em animais, principais métodos de predição e estimação de efeitos de marcadores moleculares na atualidade, a importância da simulação de dados para a validação desses métodos, bem como as vantagens, os desafios e as limitações no cenário atual da produção animal com o uso da seleção e associação genômica ampla.

Genome-enabled prediction through quantile random forest for complex traits / Predição genômica por meio do random forest quantílitico para características complexas

Quantile Random Forest (QRF) is a non-parametric methodology that combines the advantages of Random Forest (RF) and Quantile Regression (QR). Specifically, this approach can explore non-linear functions, determining the probability distribution of a response variable and extracting information from different quantiles instead of just predicting the mean. This evaluated the performance of the QRF in the genomic prediction for complex traits (epistasis and dominance). In addition, compare the accuracies obtained with those derived from the G-BLUP. The simulation created an F2 population with 1,000 individuals and genotyped for 4,010 SNP markers. Besides, twelve traits were simulated from a model considering additive and non-additive effects, QTL (Quantitative trait loci) numbers ranging from eight to 120, and heritability of 0.3, 0.5, or 0.8. For training and validation, the 5-fold cross-validation approach was used. For each fold, the accuracies of all the proposed models were calculated: QRF in five different quantiles and three G-BLUP models (additive effect, additive and epistatic effects, additive and dominant effects). Finally, the predictive performance of these methodologies was compared. In all scenarios, the QRF accuracies were equal to or greater than the methodologies evaluated and proved to be an alternative tool to predict genetic values in complex traits.

Quantile Random Forest (QRF) é uma metodologia não paramétrica, que combina as vantagens do Random Forest (RF) e da Regressão Quantílica (QR). Especificamente, essa abordagem pode explorar funções não lineares, determinando a distribuição de probabilidade de uma variável resposta e extraindo informações de diferentes quantis em vez de apenas prever a média. O objetivo deste trabalho foi avaliar o desempenho do QRF em predizer o valor genético genômico para características com arquitetura genética não aditiva (epistasia e dominância). Adicionalmente, as acurácias obtidas foram comparadas com aquelas advindas do G-BLUP. A simulação criou uma população F2 com 1.000 indivíduos genotipados para 4.010 marcadores SNP. Além disso, doze características foram simuladas a partir de um modelo considerando efeitos aditivos e não aditivos, com número de QTL (Quantitative trait loci) variando de oito a 120 e herdabilidade de 0,3, 0,5 ou 0,8. Para treinamento e validação foi usada a abordagem da validação cruzada 5-fold. Para cada um dos folds foram calculadas as acurácias de todos os modelos propostos: QRF em cinco quantis diferentes e três modelos do G-BLUP (com efeito aditivo, aditivo e epistático, aditivo e dominante). Por fim, o desempenho preditivo dessas metodologias foi comparado. Em todos os cenários, as acurácias do QRF foram iguais ou superiores às metodologias avaliadas e mostrou ser uma ferramenta alternativa para predizer valores genéticos em características complexas.

Overview of the use of genomic data in animal breeding / Panorama do uso de dados genômicos no melhoramento genético animal

The use of molecular information in breeding programs contributed to important advances in the improvement of traits of economic interest in livestock production. The advent of single nucleotide polymorphism (SNP) panels applied to genome-wide selection (GWS) and genome-wide association studies (GWAS), along with computational advances (e.g., use of powerful software and robust analyses) allowed a better understanding of the genetic architecture of farm animals and increased the selection efficiency. In this context, the statistic method single-step GBLUP has been frequently used to perform GWS, and more recently GWAS analyses, providing accurate predictions and QTL detection, respectively. Nevertheless, in developing countries, species such as sheep and goats, whose genomic data are more difficult to be obtained, the use of data simulation has been efficient in the study of the major factors involved in the selection process, such as size of training population, density of SNP chips, and genotyping strategies. The effects of these factors are directly associated with the prediction accuracy of genomic breeding values. In this review we showed important aspects of the use of genomics in the genetic improvement of production traits of animals, the main methods currently used for prediction and estimation of molecular marker effects, the importance of data simulation for validation of those methods, as well as the advantages, challenges and limitations of the use of GWS and GWAS in the current scenario of livestock production.

Em programas de melhoramento genético, o uso de informações moleculares garantiu importantes avanços para a melhoria de características de interesse econômico, no âmbito da produção animal. O advento da tecnologia de painéis de SNPs aplicados à seleção genômica ampla (GWS) e associação genômica ampla (GWAS), aliado ao avanço computacional, com o uso de softwares e análises robustas, permitiram melhor compreensão sobre a arquitetura genética dos animais de produção e, consequentemente, maior eficiência na seleção. Nesse contexto, o método estatístico single-step GBLUP tem sido utilizado, frequentemente, na execução da GWS e, mais recentemente, em GWAS, possibilitando predições acuradas e detecção de QTLs, respectivamente. No entanto, em países em desenvolvimento e, em espécies como os ovinos e caprinos, que existe maior dificuldade para a aquisição de dados genômicos, o uso da simulação de dados tem se mostrado eficiente para estudar os principais fatores envolvidos no processo de seleção, como o tamanho da população de treinamento, densidade de chipde SNPs e estratégias de genotipagem, cujos efeitos estão diretamente associados à acurácia da predição de valores genéticos genômicos. Nesta revisão, serão abordados pontos importantes sobre o uso da genômica no melhoramento genético de características produtivas em animais, principais métodos de predição e estimação de efeitos de marcadores moleculares na atualidade, a importância da simulação de dados para a validação desses métodos, bem como as vantagens, os desafios e as limitações no cenário atual da produção animal com o uso da seleção e associação genômica ampla.

Genetic parameters and association between longevity and milk production in buffaloes using the ssGBLUP method

The objectives of this work were to estimate the genetic parameters for the traits longevity (LG) and accumulated milk yield at 305 days (MY305) using a bitrait animal model and the single-step GBLUP method and estimate the genetic gain for LG through direct and indirect selection for MY305. We used 4,057 records of first lactations of Murrah dairy buffaloes, collected between 1987 and 2020, belonging to six Brazilian herds located in the states Ceará, Rio Grande do Norte, and São Paulo and 960 animals genotyped using the 90K Axiom Buffalo Genotyping (Thermo Fisher Scientific, Santa Clara, CA) to estimate the genetic parameters. The heritability estimate was 0.25 for MY305 and 0.13 for LG. The genetic gain for LG was 0.13 months under direct selection, and 0.14 months under indirect selection, which results in a relative selection efficiency of 11% under selection for MY305 compared with the direct selection. The genetic correlation between the two traits was 0.77, indicating that animals with genetic potential for high MY305 tend to live longer. The genetic trends for MY305 and LG were 0.22 kg/year and 5.20 days/year, respectively, indicating a positive response, which reaffirms its relationship with the high genetic correlation between the two traits.(AU)

Evaluation of Bayesian methods of genomic association via chromosomic regions using simulated data

The development of efficient methods for genome-wide association studies (GWAS) between quantitative trait loci (QTL) and genetic values is extremely important to animal and plant breeding programs. Bayesian approaches that aim to select regions of single nucleotide polymorphisms (SNPs) proved to be efficient, indicating genes with important effects. Among the selection criteria for SNPs or regions, selection criterion by percentage of variance can be explained by genomic regions (%var), selection of tag SNPs, and selection based on the window posterior probability of association (WPPA). To also detect potentially associated regions, we proposed measuring posterior probability of the interval PPint), which aims to select regions based on the markers of greatest effects. Therefore, the objective of this work was to evaluate these approaches, in terms of efficiency in selecting and identifying markers or regions located within or close to genes associated with traits. This study also aimed to compare these methodologies with single-marker analyses. To accomplish this, simulated data were used in six scenarios, with SNPs allocated in non-overlapping genomic regions. Considering traits with oligogenic inheritance, WPPA criterion followed by %var and PPint criteria were shown to be superior, presenting higher values of detection power, capturing higher percentages of genetic variance and larger areas. For traits with polygenic inheritance, PPint and WPPA criteria were considered superior. Single-marker analyses identified SNPs associated only in oligogenic inheritance scenarios and was lower than the other criteria.(AU)

Evaluation of a new additive-dominance genomic model and implications for quantitative genetics and genomic selection

The Fisher's infinitesimal model is traditionally used in quantitative genetics and genomic selection, and it attributes most genetic variance to additive variance. Recently, the dominance maximization model was proposed and it prioritizes the dominance variance based on alternative parameterizations. In this model, the additive effects at the locus level are introduced into the model after the dominance variance is maximized. In this study, the new parameterizations of additive and dominance effects on quantitative genetics and genomic selection were evaluated and compared with the parameterizations traditionally applied using the genomic best linear unbiased prediction method. As the parametric relative magnitude of the additive and dominance effects vary with allelic frequencies of populations, we considered different minor allele frequencies to compare the relative magnitudes. We also proposed and evaluated two indices that combine the additive and dominance variances estimated by both models. The dominance maximization model, along with the two indices, offers alternatives to improve the estimates of additive and dominance variances and their respective proportions and can be successfully used in genetic evaluation.

Determination of optimal number of independent components in yield traits in rice

The principal component regression (PCR) and the independent component regression (ICR) are dimensionality reduction methods and extremely important in genomic prediction. These methods require the choice of the number of components to be inserted into the model. For PCR, there are formal criteria; however, for ICR, the adopted criterion chooses the number of independent components (ICs) associated to greater accuracy and requires high computational time. In this study, seven criteria based on the number of principal components (PCs) and methods of variable selection to guide this choice in ICR are proposed and evaluated in simulated and real data. For both datasets, the most efficient criterion and that drastically reduced computational time determined that the number of ICs should be equal to the number of PCs to reach a higher accuracy value. In addition, the criteria did not recover the simulated heritability and generated biased genomic values.

Identification of biomarkers for bull fertility using functional genomics

Prediction of bull fertility is critical for the sustainability of both dairy and beef cattle production. Even though bulls produce ample amounts of sperm with normal parameters, some bulls may still suffer from subpar fertility. This causes major economic losses in the cattle industry because using artificial insemination, semen from one single bull can be used to inseminate hundreds of thousands of cows. Although there are several traditional methods to estimate bull fertility, such methods are not sufficient to explain and accurately predict the subfertility of individual bulls. Since fertility is a complex trait influenced by a number of factors including genetics, epigenetics, and environment, there is an urgent need for a comprehensive methodological approach to clarify uncertainty in male subfertility. The present review focuses on molecular and functional signatures of bull sperm associated with fertility. Potential roles of functional genomics (proteome, small noncoding RNAs, lipidome, metabolome) on determining male fertility and its potential as a fertility biomarker are discussed. This review provides a better understanding of the molecular signatures of viable and fertile sperm cells and their potential to be used as fertility biomarkers. This information will help uncover the underlying reasons for idiopathic subfertility.(AU)

Genomic prediction of leaf rust resistance to Arabica coffee using machine learning algorithms

Genomic selection (GS) emphasizes the simultaneous prediction of the genetic effects of thousands of scattered markers over the genome. Several statistical methodologies have been used in GS for the prediction of genetic merit. In general, such methodologies require certain assumptions about the data, such as the normality of the distribution of phenotypic values. To circumvent the non-normality of phenotypic values, the literature suggests the use of Bayesian Generalized Linear Regression (GBLASSO). Another alternative is the models based on machine learning, represented by methodologies such as Artificial Neural Networks (ANN), Decision Trees (DT) and related possible refinements such as Bagging, Random Forest and Boosting. This study aimed to use DT and its refinements for predicting resistance to orange rust in Arabica coffee. Additionally, DT and its refinements were used to identify the importance of markers related to the characteristic of interest. The results were compared with those from GBLASSO and ANN. Data on coffee rust resistance of 245 Arabica coffee plants genotyped for 137 markers were used. The DT refinements presented equal or inferior values of Apparent Error Rate compared to those obtained by DT, GBLASSO, and ANN. Moreover, DT refinements were able to identify important markers for the characteristic of interest. Out of 14 of the most important markers analyzed in each methodology, 9.3 markers on average were in regions of quantitative trait loci (QTLs) related to resistance to disease listed in the literature.

Genomic prediction of leaf rust resistance to Arabica coffee using machine learning algorithms

Genomic selection (GS) emphasizes the simultaneous prediction of the genetic effects of thousands of scattered markers over the genome. Several statistical methodologies have been used in GS for the prediction of genetic merit. In general, such methodologies require certain assumptions about the data, such as the normality of the distribution of phenotypic values. To circumvent the non-normality of phenotypic values, the literature suggests the use of Bayesian Generalized Linear Regression (GBLASSO). Another alternative is the models based on machine learning, represented by methodologies such as Artificial Neural Networks (ANN), Decision Trees (DT) and related possible refinements such as Bagging, Random Forest and Boosting. This study aimed to use DT and its refinements for predicting resistance to orange rust in Arabica coffee. Additionally, DT and its refinements were used to identify the importance of markers related to the characteristic of interest. The results were compared with those from GBLASSO and ANN. Data on coffee rust resistance of 245 Arabica coffee plants genotyped for 137 markers were used. The DT refinements presented equal or inferior values of Apparent Error Rate compared to those obtained by DT, GBLASSO, and ANN. Moreover, DT refinements were able to identify important markers for the characteristic of interest. Out of 14 of the most important markers analyzed in each methodology, 9.3 markers on average were in regions of quantitative trait loci (QTLs) related to resistance to disease listed in the literature.(AU)

New insights into genomic selection through population-based non-parametric prediction methods

Genome-wide selection (GWS) is based on a large number of markers widely distributed throughout the genome. Genome-wide selection provides for the estimation of the effect of each molecular marker on the phenotype, thereby allowing for the capture of all genes affecting the quantitative traits of interest. The main statistical tools applied to GWS are based on random regression or dimensionality reduction methods. In this study a new non-parametric method, called Delta-p was proposed, which was then compared to the Genomic Best Linear Unbiased Predictor (G-BLUP) method. Furthermore, a new selection index combining the genetic values obtained by the G-BLUP and Delta-p, named Delta-p/G-BLUP methods, was proposed. The efficiency of the proposed methods was evaluated through both simulation and real studies. The simulated data consisted of eight scenarios comprising a combination of two levels of heritability, two genetic architectures and two dominance status (absence and complete dominance). Each scenario was simulated ten times. All methods were applied to a real dataset of Asian rice (Oryza sativa) aiming to increase the efficiency of a current breeding program. The methods were compared as regards accuracy of prediction (simulation data) or predictive ability (real dataset), bias and recovery of the true genomic heritability. The results indicated that the proposed Delta-p/G-BLUP index outperformed the other methods in both prediction accuracy and predictive ability.

New insights into genomic selection through population-based non-parametric prediction methods

Genome-wide selection (GWS) is based on a large number of markers widely distributed throughout the genome. Genome-wide selection provides for the estimation of the effect of each molecular marker on the phenotype, thereby allowing for the capture of all genes affecting the quantitative traits of interest. The main statistical tools applied to GWS are based on random regression or dimensionality reduction methods. In this study a new non-parametric method, called Delta-p was proposed, which was then compared to the Genomic Best Linear Unbiased Predictor (G-BLUP) method. Furthermore, a new selection index combining the genetic values obtained by the G-BLUP and Delta-p, named Delta-p/G-BLUP methods, was proposed. The efficiency of the proposed methods was evaluated through both simulation and real studies. The simulated data consisted of eight scenarios comprising a combination of two levels of heritability, two genetic architectures and two dominance status (absence and complete dominance). Each scenario was simulated ten times. All methods were applied to a real dataset of Asian rice (Oryza sativa) aiming to increase the efficiency of a current breeding program. The methods were compared as regards accuracy of prediction (simulation data) or predictive ability (real dataset), bias and recovery of the true genomic heritability. The results indicated that the proposed Delta-p/G-BLUP index outperformed the other methods in both prediction accuracy and predictive ability.(AU)

Triple categorical regression for genomic selection: application to cassava breeding

Genome-wide selection (GWS) is currently a technique of great importance in plant breeding, since it improves efficiency of genetic evaluations by increasing genetic gains. The process is based on genomic estimated breeding values (GEBVs) obtained through phenotypic and dense marker genomic information. In this context, GEBVs of N individuals are calculated through appropriate models, which estimate the effect of each marker on phenotypes, allowing the early identification of genetically superior individuals. However, GWS leads to statistical challenges, due to high dimensionality and multicollinearity problems. These challenges require the use of statistical methods to approach the regularization of the estimation process. Therefore, we aimed to propose a method denominated as triple categorical regression (TCR) and compare it with the genomic best linear unbiased predictor (G-BLUP) and Bayesian least absolute shrinkage and selection operator (BLASSO) methods that have been widely applied to GWS. The methods were evaluated in simulated populations considering four different scenarios. Additionally, a modification of the G-BLUP method was proposed based on the TCR-estimated (TCR/G-BLUP) results. All methods were applied to real data of cassava (Manihot esculenta) with to increase efficiency of a current breeding program. The methods were compared through independent validation and efficiency measures, such as prediction accuracy, bias, and recovered genomic heritability. The TCR method was suitable to estimate variance components and heritability, and the TCR/G-BLUP method provided efficient GEBV predictions. Thus, the proposed methods provide new insights for GWS.

Triple categorical regression for genomic selection: application to cassava breeding

Genome-wide selection (GWS) is currently a technique of great importance in plant breeding, since it improves efficiency of genetic evaluations by increasing genetic gains. The process is based on genomic estimated breeding values (GEBVs) obtained through phenotypic and dense marker genomic information. In this context, GEBVs of N individuals are calculated through appropriate models, which estimate the effect of each marker on phenotypes, allowing the early identification of genetically superior individuals. However, GWS leads to statistical challenges, due to high dimensionality and multicollinearity problems. These challenges require the use of statistical methods to approach the regularization of the estimation process. Therefore, we aimed to propose a method denominated as triple categorical regression (TCR) and compare it with the genomic best linear unbiased predictor (G-BLUP) and Bayesian least absolute shrinkage and selection operator (BLASSO) methods that have been widely applied to GWS. The methods were evaluated in simulated populations considering four different scenarios. Additionally, a modification of the G-BLUP method was proposed based on the TCR-estimated (TCR/G-BLUP) results. All methods were applied to real data of cassava (Manihot esculenta) with to increase efficiency of a current breeding program. The methods were compared through independent validation and efficiency measures, such as prediction accuracy, bias, and recovered genomic heritability. The TCR method was suitable to estimate variance components and heritability, and the TCR/G-BLUP method provided efficient GEBV predictions. Thus, the proposed methods provide new insights for GWS.(AU)

Combined index of genomic prediction methods applied to productivity / Índice combinado de métodos de predição genômica aplicado em características de produtividade em arroz

Rice cultivation has great national and global importance, being one of the most produced and consumed cereals in the world and the primary food for more than half of the worlds population. Because of its importance as food, developing efficient methods to select and predict genetically superior individuals in reference to plant traits is of extreme importance for breeding programs. The objective of this research was to evaluate and compare the efficiency of the Delta-p, G-BLUP (Genomic Best Linear Unbiased Predictor), BayesCpi, BLASSO (Bayesian Least Absolute Shrinkage and Selection Operator), Delta-p/G-BLUP index, Delta-p/BayesCpi index, and Delta-p/BLASSO index in the estimation of genomic values and the effects of single nucleotide polymorphisms on phenotypic data associated with rice traits. Use of molecular markers allowed high selective efficiency and increased genetic gain per unit time. The Delta-p method uses the concept of change in allelic frequency caused by selection and the theoretical concept of genetic gain. The Index is based on the principle of combined selection, using the information regarding the additive genomic values predicted via G-BLUP, BayesCpi, BLASSO, or Delta-p. These methods were applied and compared for genomic prediction using nine rice traits: flag leaf length, flag leaf width, panicles number per plant, primary panicle branch number, seed length, seed width, amylose content, protein content, and blast resistance. Delta-p/G-BLUP index had higher predictive abilities for the traits studied, except for amylose content trait in which the method with the highest predictive ability was BayesCpi, being approximately 3% greater than that of the Delta-p/G-BLUP index.(AU)

A cultura do arroz tem grande importância nacional e mundial por ser um dos cereais mais produzidos e consumidos no mundo, caracterizando-se como o principal alimento de mais da metade da população mundial. Em função de sua importância alimentar, desenvolver métodos eficientes que visam a predição e a seleção de indivíduos geneticamente superiores, quanto a características da planta, é de extrema importância para os programas de melhoramento. Diante disso, o objetivo deste trabalho foi avaliar e comparar a eficiência do método Delta-p, G-BLUP, BayesCpi, BLASSO e o índice Delta-p/G-BLUP, índice Delta-p/BayesCpi e índice Delta-p/BLASSO, na estimação de valores genômicos e dos efeitos de marcadores SNPs (Single Nucleotide Polymorphisms) em dados fenotípicos associados a características de arroz. A utilização de marcadores moleculares permite alta eficiência seletiva e o aumento do ganho genético por unidade de tempo. O método Delta-p utiliza o conceito de mudança na frequência alélica devido à seleção e o conceito teórico de ganho genético. O Índice é baseado no princípio da seleção combinada, utiliza conjuntamente as informações dos valores genômicos aditivos preditos via G-BLUP, BayesCpi ou BLASSO e via Delta-p. Estes métodos foram aplicados e comparados quanto à predição genômica utilizando nove características de arroz (Oryza sativa), sendo elas: comprimento da folha bandeira, largura da folha bandeira; número de panículas por planta; número de ramos da panícula primária; comprimento de semente; largura de semente; teor de amilose; teor de proteína; resistência a bruzone. O índice Delta-p/G-BLUP obteve maiores capacidades preditivas para as características estudadas, exceto para a característica Conteúdo de amilose, em que o método que obteve maior capacidade preditiva foi o BayesCpi, sendo aproximadamente 3% superior ao índice Delta-p/G-BLUP.(AU)

Genome prediction accuracy of common bean via Bayesian models

ABSTRACT: We aimed to apply genomic information based on SNP (single nucleotide polymorphism) markers for the genetic evaluation of the traits stay-green (SG), plant architecture (PA), grain aspect (GA) and grain yield (GY) in common bean through Bayesian models. These models were compared in terms of prediction accuracy and ability for heritability estimation for each one of the mentioned traits. A total of 80 cultivars were genotyped for 377 SNP markers, whose effects were estimated by five different Bayesian models: Bayes A (BA), B (BB), C (BC), LASSO (BL) e Ridge regression (BRR). Although, prediction accuracies calculated by means of cross-validation have been similar within each trait, the BB model stood out for the trait SG, whereas the BRR was indicated for the remaining traits. The heritability estimates for the traits SG, PA, GA and GY were 0.61, 0.28, 0.32 and 0.29, respectively. In summary, the Bayesian methods applied here were effective and ease to be implemented. The used SNP markers can help in the early selection of promising genotypes, since incorporating genomic information increase the prediction accuracy of the estimated genetic merit.

RESUMO: Objetivou-se incorporar informações genômicas de marcadores SNP (single nucleotide polymorphism) na avaliação genética das características stay-green (SG), arquitetura de planta (AP), aspecto de grãos (AG) e produtividade de grãos (PG) em feijoeiro-comum via modelos Bayesianos. Estes modelos foram comparados quanto a acurácia de predição e habilidade de estimação da herdabilidade para cada característica. Utilizaram-se informações de 80 cultivares genotipadas para 377 marcadores SNP, cujos efeitos de substituição alélica foram estimados por meio de cinco diferentes modelos Bayesianos: Bayes A (BA), B (BB), C (BC), LASSO (BL) e regressão ridge (BRR). Embora as acurácias de predição calculadas por meio de análise de validação cruzada tenham sido similares dentro de cada característica, o modelo BB se destacou para a característica SG, enquanto o modelo BRR foi indicado para as demais. As herdabilidades estimadas para SG, AP, AG e PG foram, respectivamente, 0,61, 0,28, 0,32 e 0,29. Em resumo, os métodos contemplados mostraram-se efetivos e de fácil implementação. O conjunto de marcadores utilizado pode auxiliar na seleção precoce de genótipos promissores, uma vez que a incorporação de informações genômicas aumenta a acurácia de predição do mérito genético estimado.

Evaluation of imputed genomic data in discrete traits using Random forest and Bayesian threshold methods / Avaliação de dados genômicos imputados em características distintas usando os métodos de Random Forest e de limiares Bayesianos

The objectives of this study were (1) to quantify imputation accuracy and to assess the factors affecting it; and (2) to evaluate the accuracy of threshold BayesA (TBA), Bayesian threshold LASSO (BTL) and random forest (RF) algorithms to analyze discrete traits. Genomic data were simulated to reflect variations in heritability (h2 = 0.30 and 0.10), number of QTL (QTL = 81 and 810), number of SNP (10 K and 50 K) and linkage disequilibrium (LD=low and high) for 27 chromosomes. For real condition simulating, we randomly masked markers with 90% missing rate for each scenario; afterwards, hidden markers were imputed using FImpute software. In imputed genotypes, a wide range of accuracy was observed for RF (0.164-0.512) compared to TBA (0.283-0.469) and BTL (0.272-0.504). Comparing to original genotypes, using imputed genotypes decreased the average accuracy of genomic prediction about 0.0273 (range of 0.024 to 0.036). Comparing to Bayesian threshold, using RF was improved rapidly accuracy of genomic prediction with increase in the marker density. Despite the higher accuracy of BTL and TBA at different levels of LD and heritability, the increase in accuracy was greater for RF. Furthermore, the best method for prediction of genomic accuracy depends on genomic architecture of population.(AU)

Os objetivos deste estudo foram (1) quantificar a precisão de imputação e acessar os fatores que as afetam; e (2) avaliar a precisão do princípio de BayesA (TBA), do modelo Bayesiano LASSO (BTL), e o algoritmo Random Forest para analisar as características distintas. Dados genômicos foram simulados para indicar variações na herdabilidade (h2 = 0.30 e 0.10), número de QTL (QTL = 81 e 810), número de SNP (10 k e 50 k) e desequilíbrio de ligação (LD = baixo e alto) para 27 cromossomos. Para uma simulação mais realista, nós cobrimos os marcadores aleatoriamente com 90% da taxa ausente para cada cenário, depois, os marcadores foram imputados usando o software FImpute. Nos genótipos imputados uma grande oscilação de precisão foi observada pelo modelo RF (0.164-0.512) comparado com TBA (0.283 - 0.469) e BTL (0.272 - 0.504). Comparando com os genótipos originais, os genótipos imputados decaíram a precisão média da predição genômica em cerca de 0.0273 (oscilação de 0.024 para 0.036). Comparando-se ao princípio Bayesiano, o uso de RF melhorou a precisão de predição com o aumento da densidade do marcador. Além disso, o melhor método para predição de precisão genômica depende da arquitetura genômica da sua população.(AU)

Evaluation of imputed genomic data in discrete traits using Random forest and Bayesian threshold methods / Avaliação de dados genômicos imputados em características distintas usando os métodos de Random Forest e de limiares Bayesianos

The objectives of this study were (1) to quantify imputation accuracy and to assess the factors affecting it; and (2) to evaluate the accuracy of threshold BayesA (TBA), Bayesian threshold LASSO (BTL) and random forest (RF) algorithms to analyze discrete traits. Genomic data were simulated to reflect variations in heritability (h2 = 0.30 and 0.10), number of QTL (QTL = 81 and 810), number of SNP (10 K and 50 K) and linkage disequilibrium (LD=low and high) for 27 chromosomes. For real condition simulating, we randomly masked markers with 90% missing rate for each scenario; afterwards, hidden markers were imputed using FImpute software. In imputed genotypes, a wide range of accuracy was observed for RF (0.164-0.512) compared to TBA (0.283-0.469) and BTL (0.272-0.504). Comparing to original genotypes, using imputed genotypes decreased the average accuracy of genomic prediction about 0.0273 (range of 0.024 to 0.036). Comparing to Bayesian threshold, using RF was improved rapidly accuracy of genomic prediction with increase in the marker density. Despite the higher accuracy of BTL and TBA at different levels of LD and heritability, the increase in accuracy was greater for RF. Furthermore, the best method for prediction of genomic accuracy depends on genomic architecture of population.

Os objetivos deste estudo foram (1) quantificar a precisão de imputação e acessar os fatores que as afetam; e (2) avaliar a precisão do princípio de BayesA (TBA), do modelo Bayesiano LASSO (BTL), e o algoritmo Random Forest para analisar as características distintas. Dados genômicos foram simulados para indicar variações na herdabilidade (h2 = 0.30 e 0.10), número de QTL (QTL = 81 e 810), número de SNP (10 k e 50 k) e desequilíbrio de ligação (LD = baixo e alto) para 27 cromossomos. Para uma simulação mais realista, nós cobrimos os marcadores aleatoriamente com 90% da taxa ausente para cada cenário, depois, os marcadores foram imputados usando o software FImpute. Nos genótipos imputados uma grande oscilação de precisão foi observada pelo modelo RF (0.164-0.512) comparado com TBA (0.283 - 0.469) e BTL (0.272 - 0.504). Comparando com os genótipos originais, os genótipos imputados decaíram a precisão média da predição genômica em cerca de 0.0273 (oscilação de 0.024 para 0.036). Comparando-se ao princípio Bayesiano, o uso de RF melhorou a precisão de predição com o aumento da densidade do marcador. Além disso, o melhor método para predição de precisão genômica depende da arquitetura genômica da sua população.

Genome prediction accuracy of common bean via Bayesian models / Acurácia de predição genômica em feijoeiro-comum via modelos Bayesianos

We aimed to apply genomic information based on SNP (single nucleotide polymorphism) markers for the genetic evaluation of the traits stay-green (SG), plant architecture (PA), grain aspect (GA) and grain yield (GY) in common bean through Bayesian models. These models were compared in terms of prediction accuracy and ability for heritability estimation for each one of the mentioned traits. A total of 80 cultivars were genotyped for 377 SNP markers, whose effects were estimated by five different Bayesian models: Bayes A (BA), B (BB), C (BC), LASSO (BL) e Ridge regression (BRR). Although, prediction accuracies calculated by means of cross-validation have been similar within each trait, the BB model stood out for the trait SG, whereas the BRR was indicated for the remaining traits. The heritability estimates for the traits SG, PA, GA and GY were 0.61, 0.28, 0.32 and 0.29, respectively. In summary, the Bayesian methods applied here were effective and ease to be implemented. The used SNP markers can help in the early selection of promising genotypes, since incorporating genomic information increase the prediction accuracy of the estimated genetic merit.(AU)

Objetivou-se incorporar informações genômicas de marcadores SNP (single nucleotide polymorphism) na avaliação genética das características stay-green (SG), arquitetura de planta (AP), aspecto de grãos (AG) e produtividade de grãos (PG) em feijoeiro-comum via modelos Bayesianos. Estes modelos foram comparados quanto a acurácia de predição e habilidade de estimação da herdabilidade para cada característica. Utilizaram-se informações de 80 cultivares genotipadas para 377 marcadores SNP, cujos efeitos de substituição alélica foram estimados por meio de cinco diferentes modelos Bayesianos: Bayes A (BA), B (BB), C (BC), LASSO (BL) e regressão ridge (BRR). Embora as acurácias de predição calculadas por meio de análise de validação cruzada tenham sido similares dentro de cada característica, o modelo BB se destacou para a característica SG, enquanto o modelo BRR foi indicado para as demais. As herdabilidades estimadas para SG, AP, AG e PG foram, respectivamente, 0,61, 0,28, 0,32 e 0,29. Em resumo, os métodos contemplados mostraram-se efetivos e de fácil implementação. O conjunto de marcadores utilizado pode auxiliar na seleção precoce de genótipos promissores, uma vez que a incorporação de informações genômicas aumenta a acurácia de predição do mérito genético estimado.(AU)