Search

In sorghum [Sorghum bicolor (L.) Moench], hybrid cultivars for the biofuel industry are desired. Along with selection based on testcross performance, evaluation of the breeding population per se is also important for the success of hybrid breeding. In addition to additive genetic effects, non-additive (i.e., dominance and epistatic) effects are expected to contribute to the performance of early generations. Unfortunately, studies on early generations in sorghum breeding programs are limited. In this study, we analyzed a breeding population for bioenergy sorghum, which was previously developed based on testcross performance, to compare genomic selection models both trained on and evaluated for the per se performance of the 3rd generation S0 individuals. Of over 200 ancestral inbred accessions in the base population, only 13 founders contributed to the 3rd generation as progenitors. Compared to the founders, the performances of the population per se were improved for target traits. The total genetic variance within the S0 generation progenies themselves for all traits was mainly additive, although non-additive variances contributed to each trait to some extent. For genomic selection, linear regression models explicitly considering all genetic components showed a higher predictive ability than other linear and non-linear models. Although the number and effect distribution of underlying loci was different among the traits, the influence of priors for marker effects was relatively small. These results indicate the importance of considering non-additive effects for dissecting the genetic architecture of early breeding generations and predicting the performance per se.

The paradigm called genomic selection (GS) is a revolutionary way of developing new plants and animals. This is a predictive methodology, since it uses learning methods to perform its task. Unfortunately, there is no universal model that can be used for all types of predictions; for this reason, specific methodologies are required for each type of output (response variables). Since there is a lack of efficient methodologies for multivariate count data outcomes, in this paper, a multivariate Poisson deep neural network (MPDN) model is proposed for the genomic prediction of various count outcomes simultaneously. The MPDN model uses the minus log-likelihood of a Poisson distribution as a loss function, in hidden layers for capturing nonlinear patterns using the rectified linear unit (RELU) activation function and, in the output layer, the exponential activation function was used for producing outputs on the same scale of counts. The proposed MPDN model was compared to conventional generalized Poisson regression models and univariate Poisson deep learning models in two experimental data sets of count data. We found that the proposed MPDL outperformed univariate Poisson deep neural network models, but did not outperform, in terms of prediction, the univariate generalized Poisson regression models. All deep learning models were implemented in Tensorflow as back-end and Keras as front-end, which allows implementing these models on moderate and large data sets, which is a significant advantage over previous GS models for multivariate count data.

Due to the ever-increasing data collected in genomic breeding programs, there is a need for genomic prediction models that can deal better with big data. For this reason, here we propose a Maximum a posteriori Threshold Genomic Prediction (MAPT) model for ordinal traits that is more efficient than the conventional Bayesian Threshold Genomic Prediction model for ordinal traits. The MAPT performs the predictions of the Threshold Genomic Prediction model by using the maximum a posteriori estimation of the parameters, that is, the values of the parameters that maximize the joint posterior density. We compared the prediction performance of the proposed MAPT to the conventional Bayesian Threshold Genomic Prediction model, the multinomial Ridge regression and support vector machine on 8 real data sets. We found that the proposed MAPT was competitive with regard to the multinomial and support vector machine models in terms of prediction performance, and slightly better than the conventional Bayesian Threshold Genomic Prediction model. With regard to the implementation time, we found that in general the MAPT and the support vector machine were the best, while the slowest was the multinomial Ridge regression model. However, it is important to point out that the successful implementation of the proposed MAPT model depends on the informative priors used to avoid underestimation of variance components.

Improving fruit quality is an important but challenging breeding goal in winter squash. Squash breeding in general is resource-intensive, especially in terms of space, and the biology of squash makes it difficult to practice selection on both parents. These restrictions translate to smaller breeding populations and limited use of greenhouse generations, which in turn, limit genetic gain per breeding cycle and increases cycle length. Genomic selection is a promising technology for improving breeding efficiency; yet, few studies have explored its use in horticultural crops. We present results demonstrating the predictive ability of whole-genome models for fruit quality traits. Predictive abilities for quality traits were low to moderate, but sufficient for implementation. To test the use of genomic selection for improving fruit quality, we conducted three rounds of genomic recurrent selection in a butternut squash (Cucurbita moschata) population. Selections were based on a fruit quality index derived from a multi-trait genomic selection model. Remnant seed from selected populations was used to assess realized gain from selection. Analysis revealed significant improvement in fruit quality index value and changes in correlated traits. This study is one of the first empirical studies to evaluate gain from a multi-trait genomic selection model in a resource-limited horticultural crop.

Most of the genomic studies in plants and animals have used additive models for studying genetic parameters and prediction accuracies. In this study, we used genomic models with additive and nonadditive effects to analyze the genetic architecture of growth and wood traits in an open-pollinated (OP) population of Eucalyptus pellita. We used two progeny trials consisting of 5742 trees from 244 OP families to estimate genetic parameters and to test genomic prediction accuracies of three growth traits (diameter at breast height - DBH, total height - Ht and tree volume - Vol) and kraft pulp yield (KPY). From 5742 trees, 468 trees from 28 families were genotyped with 2023 pre-selected markers from candidate genes. We used the pedigree-based additive best linear unbiased prediction (ABLUP) model and two marker-based models (single-step genomic BLUP – ssGBLUP and genomic BLUP – GBLUP) to estimate the genetic parameters and compare the prediction accuracies. Analyses with the two genomic models revealed large dominant effects influencing the growth traits but not KPY. Theoretical breeding value accuracies were higher with the dominance effect in ssGBLUP model for the three growth traits. Accuracies of cross-validation with random folding in the genotyped trees have ranged from 0.60 to 0.82 in different models. Accuracies of ABLUP were lower than the genomic models. Accuracies ranging from 0.50 to 0.76 were observed for within family cross-validation predictions with low relationships between training and validation populations indicating part of the functional variation is captured by the markers through short-range linkage disequilibrium (LD). Within-family phenotype predictive abilities and prediction accuracies of genetic values with dominance effects are higher than the additive models for growth traits indicating the importance of dominance effects in predicting phenotypes and genetic values. This study demonstrates the importance of genomic approaches in OP families to study nonadditive effects. To capture the LD between markers and the quantitative trait loci (QTL) it may be important to use informative markers from candidate genes.

Husk has multiple functions such as protecting ears from diseases, infection, and dehydration during development. Additionally, husks comprised of fewer, shorter, thinner, and narrower layers allow faster moisture evaporation of kernels prior to harvest. Intensive studies have been conducted to identify appropriate husk architecture by understanding the genetic basis of related traits, including husk length, husk layer number, husk thickness, and husk width. However, marker-assisted selection is inefficient because the identified quantitative trait loci and associated genetic loci could only explain a small proportion of total phenotypic variation. Genomic selection (GS) has been used successfully on many species including maize on other traits. Thus, the potential of using GS for husk traits to directly identify superior inbred lines, without knowing the specific underlying genetic loci, is well worth exploring. In this study, we compared four GS models on a maize association population with 498 inbred lines belonging to four subpopulations, including 27 lines in stiff stalk, 67 lines in non-stiff stalk, 193 lines in tropical-subtropical, and 211 lines in mixture subpopulations. Genomic Best Linear Unbiased Prediction with principal components as cofactor, performed the best and was selected to examine the impact of interaction between sampling proportions and subpopulations. We found that predictions on inbred lines in a subpopulation were benefited from excluding individuals from other subpopulations for training if the training population within the subpopulation was large enough. Husk thickness exhibited the highest prediction accuracy among all husk traits. These results gave strategic insight to improve husk architecture.

Zinc (Zn) deficiency is a major risk factor for human health, affecting about 30% of the world’s population. To study the potential of genomic selection (GS) for maize with increased Zn concentration, an association panel and two doubled haploid (DH) populations were evaluated in three environments. Three genomic prediction models, M (M1: Environment + Line, M2: Environment + Line + Genomic, and M3: Environment + Line + Genomic + Genomic x Environment) incorporating main effects (lines and genomic) and the interaction between genomic and environment (G x E) were assessed to estimate the prediction ability (rMP) for each model. Two distinct cross-validation (CV) schemes simulating two genomic prediction breeding scenarios were used. CV1 predicts the performance of newly developed lines, whereas CV2 predicts the performance of lines tested in sparse multi-location trials. Predictions for Zn in CV1 ranged from −0.01 to 0.56 for DH1, 0.04 to 0.50 for DH2 and −0.001 to 0.47 for the association panel. For CV2, rMP values ranged from 0.67 to 0.71 for DH1, 0.40 to 0.56 for DH2 and 0.64 to 0.72 for the association panel. The genomic prediction model which included G x E had the highest average rMP for both CV1 (0.39 and 0.44) and CV2 (0.71 and 0.51) for the association panel and DH2 population, respectively. These results suggest that GS has potential to accelerate breeding for enhanced kernel Zn concentration by facilitating selection of superior genotypes.

Miscanthus is a perennial grass with potential for lignocellulosic ethanol production. To ensure its utility for this purpose, breeding efforts should focus on increasing genetic diversity of the nothospecies Miscanthus × giganteus (M×g) beyond the single clone used in many programs. Germplasm from the corresponding parental species M. sinensis (Msi) and M. sacchariflorus (Msa) could theoretically be used as training sets for genomic prediction of M×g clones with optimal genomic estimated breeding values for biofuel traits. To this end, we first showed that subpopulation structure makes a substantial contribution to the genomic selection (GS) prediction accuracies within a 538-member diversity panel of predominately Msi individuals and a 598-member diversity panels of Msa individuals. We then assessed the ability of these two diversity panels to train GS models that predict breeding values in an interspecific diploid 216-member M×g F2 panel. Low and negative prediction accuracies were observed when various subsets of the two diversity panels were used to train these GS models. To overcome the drawback of having only one interspecific M×g F2 panel available, we also evaluated prediction accuracies for traits simulated in 50 simulated interspecific M×g F2 panels derived from different sets of Msi and diploid Msa parents. The results revealed that genetic architectures with common causal mutations across Msi and Msa yielded the highest prediction accuracies. Ultimately, these results suggest that the ideal training set should contain the same causal mutations segregating within interspecific M×g populations, and thus efforts should be undertaken to ensure that individuals in the training and validation sets are as closely related as possible.

Modern crop breeding is in constant demand for new genetic diversity as part of the arms race with genetic gain. The elite gene pool has limited genetic variation and breeders are trying to introduce novelty from unadapted germplasm, landraces and wild relatives. For polygenic traits, currently available approaches to introgression are not ideal, as there is a demonstrable bias against exotic alleles during selection. Here, we propose a partitioned form of genomic selection, called Origin Specific Genomic Selection (OSGS), where we identify and target selection on favorable exotic alleles. Briefly, within a population derived from a bi-parental cross, we isolate alleles originating from the elite and exotic parents, which then allows us to separate out the predicted marker effects based on the allele origins. We validated the usefulness of OSGS using two nested association mapping (NAM) datasets: barley NAM (elite-exotic) and maize NAM (elite-elite), as well as by computer simulation. Our results suggest that OSGS works well in its goal to increase the contribution of favorable exotic alleles in bi-parental crosses, and it is possible to extend the approach to broader multi-parental populations.

The use of haplotypes may improve the accuracy of genomic prediction over single SNPs because haplotypes can better capture linkage disequilibrium and genomic similarity in different lines and may capture local high-order allelic interactions. Additionally, prediction accuracy could be improved by portraying population structure in the calibration set. A set of 383 advanced lines and cultivars that represent the diversity of the University of Minnesota wheat breeding program was phenotyped for yield, test weight, and protein content and genotyped using the Illumina 90K SNP Assay. Population structure was confirmed using single SNPs. Haplotype blocks of 5, 10, 15, and 20 adjacent markers were constructed for all chromosomes. A multi-allelic haplotype prediction algorithm was implemented and compared with single SNPs using both k-fold cross validation and stratified sampling optimization. After confirming population structure, the stratified sampling improved the predictive ability compared with k-fold cross validation for yield and protein content, but reduced the predictive ability for test weight. In all cases, haplotype predictions outperformed single SNPs. Haplotypes of 15 adjacent markers showed the best improvement in accuracy for all traits; however, this was more pronounced in yield and protein content. The combined use of haplotypes of 15 adjacent markers and training population optimization significantly improved the predictive ability for yield and protein content by 14.3 (four percentage points) and 16.8% (seven percentage points), respectively, compared with using single SNPs and k-fold cross validation. These results emphasize the effectiveness of using haplotypes in genomic selection to increase genetic gain in self-fertilized crops.

“Sparse testing” refers to reduced multi-environment breeding trials in which not all genotypes of interest are grown in each environment. Using genomic-enabled prediction and a model embracing genotype × environment interaction (GE), the non-observed genotype-in-environment combinations can be predicted. Consequently, the overall costs can be reduced and the testing capacities can be increased. The accuracy of predicting the unobserved data depends on different factors including (1) how many genotypes overlap between environments, (2) in how many environments each genotype is grown, and (3) which prediction method is used. In this research, we studied the predictive ability obtained when using a fixed number of plots and different sparse testing designs. The considered designs included the extreme cases of (1) no overlap of genotypes between environments, and (2) complete overlap of the genotypes between environments. In the latter case, the prediction set fully consists of genotypes that have not been tested at all. Moreover, we gradually go from one extreme to the other considering (3) intermediates between the two previous cases with varying numbers of different or non-overlapping (NO)/overlapping (O) genotypes. The empirical study is built upon two different maize hybrid data sets consisting of different genotypes crossed to two different testers (T1 and T2) and each data set was analyzed separately. For each set, phenotypic records on yield from three different environments are available. Three different prediction models were implemented, two main effects models (M1 and M2), and a model (M3) including GE. The results showed that the genome-based model including GE (M3) captured more phenotypic variation than the models that did not include this component. Also, M3 provided higher prediction accuracy than models M1 and M2 for the different allocation scenarios. Reducing the size of the calibration sets decreased the prediction accuracy under all allocation designs with M3 being the less affected model; however, using the genome-enabled models (i.e., M2 and M3) the predictive ability is recovered when more genotypes are tested across environments. Our results indicate that a substantial part of the testing resources can be saved when using genome-based models including GE for optimizing sparse testing designs.

Plant breeders regularly evaluate multiple traits across multiple environments, which opens an avenue for using multiple traits in genomic prediction models. We assessed the potential of multi-trait (MT) genomic prediction model through evaluating several strategies of incorporating multiple traits (eight agronomic and malting quality traits) into the prediction models with two cross-validation schemes (CV1, predicting new lines with genotypic information only and CV2, predicting partially phenotyped lines using both genotypic and phenotypic information from correlated traits) in barley. The predictive ability was similar for single (ST-CV1) and multi-trait (MT-CV1) models to predict new lines. However, the predictive ability for agronomic traits was considerably increased when partially phenotyped lines (MT-CV2) were used. The predictive ability for grain yield using the MT-CV2 model with other agronomic traits resulted in 57% and 61% higher predictive ability than ST-CV1 and MT-CV1 models, respectively. Therefore, complex traits such as grain yield are better predicted when correlated traits are used. Similarly, a considerable increase in the predictive ability of malting quality traits was observed when correlated traits were used. The predictive ability for grain protein content using the MT-CV2 model with both agronomic and malting traits resulted in a 76% higher predictive ability than ST-CV1 and MT-CV1 models. Additionally, the higher predictive ability for new environments was obtained for all traits using the MT-CV2 model compared to the MT-CV1 model. This study showed the potential of improving the genomic prediction of complex traits by incorporating the information from multiple traits (cost-friendly and easy to measure traits) collected throughout breeding programs which could assist in speeding up breeding cycles.

Many breeds of modern cattle are naturally horned, and for sound husbandry management reasons the calves frequently undergo procedures to physically remove the horns by disbudding or dehorning. These procedures are however a welfare concern. Selective breeding for polledness – absence of horns – has been effective in some cattle breeds but not in others (Bos indicus genotypes) due in part to the complex genetics of horn phenotype. To address this problem different approaches to genetic testing which provide accurate early-in-life prediction of horn phenotype have been evaluated, initially using microsatellites (MSAT) and more recently single nucleotide polymorphism (SNP). A direct gene test is not effective given the genetic heterogeneity and large-sized sequence variants associated with polledness in different breeds. The current study investigated 39,943 animals of multiple breeds to assess the accuracy of available poll testing assays. While the standard SNP-based test was an improvement on the earlier MSAT haplotyping method, 1999 (9.69%) out of 20,636 animals tested with this SNP-based assay did not predict a genotype, most commonly associated with the Indicus-influenced breeds. The current study has developed an optimized poll gene test that resolved the vast majority of these 1999 unresolved animals, while the predicted genotypes of those previously resolved remained unchanged. Hence the optimized poll test successfully predicted a genotype in 99.96% of samples assessed. We demonstrated that a robust set of 5 SNPs can effectively determine PC and PF alleles and eliminate the ambiguous and undetermined results of poll gene testing previously identified as an issue in cattle.

Prediction of phenotypes from genotypes is an important objective to fulfill the promises of genomics, precision medicine and agriculture. Although it’s now possible to account for the majority of genetic variation through model fitting, prediction of phenotypes remains a challenge, especially across populations that have diverged in the past. In this study, we designed simulation experiments to specifically investigate the role of genetic interactions in failure of polygenic prediction. We found that non-additive genetic interactions can significantly reduce the accuracy of polygenic prediction. Our study demonstrated the importance of considering genetic interactions in genetic prediction.

Recent literature suggests machine learning methods can capture interactions between loci and therefore could outperform linear models when predicting traits with relevant epistatic effects. However, investigating this empirically requires data with high mapping resolution and phenotypes for traits with known non-additive gene action. The objective of the present study was to compare the performance of linear (GBLUP, BayesB and elastic net [ENET]) methods to a non-parametric tree-based ensemble (gradient boosting machine – GBM) method for genomic prediction of complex traits in mice. The dataset used contained phenotypic and genotypic information for 835 animals from 6 non-overlapping generations. Traits analyzed were bone mineral density (BMD), body weight at 10, 15 and 20 weeks (BW10, BW15 and BW20), fat percentage (FAT%), circulating cholesterol (CHOL), glucose (GLUC), insulin (INS) and triglycerides (TGL), and urine creatinine (UCRT). After quality control, the genotype dataset contained 50,112 SNP markers. Animals from older generations were considered as a reference subset, while animals in the latest generation as candidates for the validation subset. We also evaluated the impact of different levels of connectedness between reference and validation sets. Model performance was measured as the Pearson’s correlation coefficient and mean squared error (MSE) between adjusted phenotypes and the model’s prediction for animals in the validation subset. Outcomes were also compared across models by checking the overlapping top markers and animals. Linear models outperformed GBM for seven out of ten traits. For these models, accuracy was proportional to the trait’s heritability. For traits BMD, CHOL and GLU, the GBM model showed better prediction accuracy and lower MSE. Interestingly, for these three traits there is evidence in literature of a relevant portion of phenotypic variance being explained by epistatic effects. We noticed that for lower connectedness, i.e., imposing a gap of one to two generations between reference and validation populations, the superior performance of GBM was only maintained for GLU. Using a subset of top markers selected from a GBM model helped for some of the traits to improve accuracy of prediction when these were fitted into linear and GBM models. The GBM model showed consistently fewer markers and animals in common among the top ranked than linear models. Our results indicate that GBM is more strongly affected by data size and decreased connectedness between reference and validation sets than the linear models. Nevertheless, our results indicate that GBM is a competitive method to predict complex traits in an outbred mice population, especially for traits with assumed epistatic effects.

The ability to connect genetic information between traits over time allow Bayesian networks to offer a powerful probabilistic framework to construct genomic prediction models. In this study, we phenotyped a diversity panel of 869 biomass sorghum (Sorghum bicolor (L.) Moench) lines, which had been genotyped with 100,435 SNP markers, for plant height (PH) with biweekly measurements from 30 to 120 days after planting (DAP) and for end-of-season dry biomass yield (DBY) in four environments. We evaluated five genomic prediction models: Bayesian network (BN), Pleiotropic Bayesian network (PBN), Dynamic Bayesian network (DBN), multi-trait GBLUP (MTr-GBLUP), and multi-time GBLUP (MTi-GBLUP) models. In fivefold cross-validation, prediction accuracies ranged from 0.46 (PBN) to 0.49 (MTr-GBLUP) for DBY and from 0.47 (DBN, DAP120) to 0.75 (MTi-GBLUP, DAP60) for PH. Forward-chaining cross-validation further improved prediction accuracies of the DBN, MTi-GBLUP and MTr-GBLUP models for PH (training slice: 30-45 DAP) by 36.4–52.4% relative to the BN and PBN models. Coincidence indices (target: biomass, secondary: PH) and a coincidence index based on lines (PH time series) showed that the ranking of lines by PH changed minimally after 45 DAP. These results suggest a two-level indirect selection method for PH at harvest (first-level target trait) and DBY (second-level target trait) could be conducted earlier in the season based on ranking of lines by PH at 45 DAP (secondary trait). With the advance of high-throughput phenotyping technologies, our proposed two-level indirect selection framework could be valuable for enhancing genetic gain per unit of time when selecting on developmental traits.

The constrained linear genomic selection index (CLGSI) is a linear combination of genomic estimated breeding values useful for predicting the net genetic merit, which in turn is a linear combination of true unobservable breeding values of the traits weighted by their respective economic values. The CLGSI is the most general genomic index and allows imposing constraints on the expected genetic gain per trait to make some traits change their mean values based on a predetermined level, while the rest of them remain without restrictions. In addition, it includes the unconstrained linear genomic index as a particular case. Using two real datasets and simulated data for seven selection cycles, we compared the theoretical results of the CLGSI with the theoretical results of the constrained linear phenotypic selection index (CLPSI). The criteria used to compare CLGSI vs. CLPSI efficiency were the estimated expected genetic gain per trait values, the selection response, and the interval between selection cycles. The results indicated that because the interval between selection cycles is shorter for the CLGSI than for the CLPSI, CLGSI is more efficient than CLPSI per unit of time, but its efficiency could be lower per selection cycle. Thus, CLGSI is a good option for performing genomic selection when there are genotyped candidates for selection.

Implementing genomic-based prediction models in genomic selection requires an understanding of the measures for evaluating prediction accuracy from different models and methods using multi-trait data. In this study, we compared prediction accuracy using six large multi-trait wheat data sets (quality and grain yield). The data were used to predict 1 year (testing) from the previous year (training) to assess prediction accuracy using four different prediction models. The results indicated that the conventional Pearson’s correlation between observed and predicted values underestimated the true correlation value, whereas the corrected Pearson’s correlation calculated by fitting a bivariate model was higher than the division of the Pearson’s correlation by the squared root of the heritability across traits, by 2.53–11.46%. Across the datasets, the corrected Pearson’s correlation was higher than the uncorrected by 5.80–14.01%. Overall, we found that for grain yield the prediction performance was highest using a multi-trait compared to a single-trait model. The higher the absolute genetic correlation between traits the greater the benefits of multi-trait models for increasing the genomic-enabled prediction accuracy of traits.

The marker density, the heritability level of trait and the statistical models adopted are critical to the accuracy of genomic prediction (GP) or selection (GS). If the potential of GP is to be fully utilized to optimize the effect of breeding and selection, in addition to incorporating the above factors into simulated data for analysis, it is essential to incorporate these factors into real data for understanding their impact on GP accuracy, more clearly and intuitively. Herein, we studied the GP of six wool traits of sheep by two different models, including Bayesian Alphabet (BayesA, BayesB, BayesCπ, and Bayesian LASSO) and genomic best linear unbiased prediction (GBLUP). We adopted fivefold cross-validation to perform the accuracy evaluation based on the genotyping data of Alpine Merino sheep (n = 821). The main aim was to study the influence and interaction of different models and marker densities on GP accuracy. The GP accuracy of the six traits was found to be between 0.28 and 0.60, as demonstrated by the cross-validation results. We showed that the accuracy of GP could be improved by increasing the marker density, which is closely related to the model adopted and the heritability level of the trait. Moreover, based on two different marker densities, it was derived that the prediction effect of GBLUP model for traits with low heritability was better; while with the increase of heritability level, the advantage of Bayesian Alphabet would be more obvious, therefore, different models of GP are appropriate in different traits. These findings indicated the significance of applying appropriate models for GP which would assist in further exploring the optimization of GP.

The usefulness of genomic prediction in crop and livestock breeding programs has prompted efforts to develop new and improved genomic prediction algorithms, such as artificial neural networks and gradient tree boosting. However, the performance of these algorithms has not been compared in a systematic manner using a wide range of datasets and models. Using data of 18 traits across six plant species with different marker densities and training population sizes, we compared the performance of six linear and six non-linear algorithms. First, we found that hyperparameter selection was necessary for all non-linear algorithms and that feature selection prior to model training was critical for artificial neural networks when the markers greatly outnumbered the number of training lines. Across all species and trait combinations, no one algorithm performed best, however predictions based on a combination of results from multiple algorithms (i.e., ensemble predictions) performed consistently well. While linear and non-linear algorithms performed best for a similar number of traits, the performance of non-linear algorithms vary more between traits. Although artificial neural networks did not perform best for any trait, we identified strategies (i.e., feature selection, seeded starting weights) that boosted their performance to near the level of other algorithms. Our results highlight the importance of algorithm selection for the prediction of trait values.

Kernel methods are flexible and easy to interpret and have been successfully used in genomic-enabled prediction of various plant species. Kernel methods used in genomic prediction comprise the linear genomic best linear unbiased predictor (GBLUP or GB) kernel, and the Gaussian kernel (GK). In general, these kernels have been used with two statistical models: single-environment and genomic × environment (GE) models. Recently near infrared spectroscopy (NIR) has been used as an inexpensive and non-destructive high-throughput phenotyping method for predicting unobserved line performance in plant breeding trials. In this study, we used a non-linear arc-cosine kernel (AK) that emulates deep learning artificial neural networks. We compared AK prediction accuracy with the prediction accuracy of GB and GK kernel methods in four genomic data sets, one of which also includes pedigree and NIR information. Results show that for all four data sets, AK and GK kernels achieved higher prediction accuracy than the linear GB kernel for the single-environment and GE multi-environment models. In addition, AK achieved similar or slightly higher prediction accuracy than the GK kernel. For all data sets, the GE model achieved higher prediction accuracy than the single-environment model. For the data set that includes pedigree, markers and NIR, results show that the NIR wavelength alone achieved lower prediction accuracy than the genomic information alone; however, the pedigree plus NIR information achieved only slightly lower prediction accuracy than the marker plus the NIR high-throughput data.

Incorporating measurements on correlated traits into genomic prediction models can increase prediction accuracy and selection gain. However, multi-trait genomic prediction models are complex and prone to overfitting which may result in a loss of prediction accuracy relative to single-trait genomic prediction. Cross-validation is considered the gold standard method for selecting and tuning models for genomic prediction in both plant and animal breeding. When used appropriately, cross-validation gives an accurate estimate of the prediction accuracy of a genomic prediction model, and can effectively choose among disparate models based on their expected performance in real data. However, we show that a naive cross-validation strategy applied to the multi-trait prediction problem can be severely biased and lead to sub-optimal choices between single and multi-trait models when secondary traits are used to aid in the prediction of focal traits and these secondary traits are measured on the individuals to be tested. We use simulations to demonstrate the extent of the problem and propose three partial solutions: 1) a parametric solution from selection index theory, 2) a semi-parametric method for correcting the cross-validation estimates of prediction accuracy, and 3) a fully non-parametric method which we call CV2*: validating model predictions against focal trait measurements from genetically related individuals. The current excitement over high-throughput phenotyping suggests that more comprehensive phenotype measurements will be useful for accelerating breeding programs. Using an appropriate cross-validation strategy should more reliably determine if and when combining information across multiple traits is useful.

Fillet yield (FY) and harvest weight (HW) are economically important traits in Nile tilapia production. Genetic improvement of these traits, especially for FY, are lacking, due to the absence of efficient methods to measure the traits without sacrificing fish and the use of information from relatives to selection. However, genomic information could be used by genomic selection to improve traits that are difficult to measure directly in selection candidates, as in the case of FY. The objectives of this study were: (i) to perform genome-wide association studies (GWAS) to dissect the genetic architecture of FY and HW, (ii) to evaluate the accuracy of genotype imputation and (iii) to assess the accuracy of genomic selection using true and imputed low-density (LD) single nucleotide polymorphism (SNP) panels to determine a cost-effective strategy for practical implementation of genomic information in tilapia breeding programs. The data set consisted of 5,866 phenotyped animals and 1,238 genotyped animals (108 parents and 1,130 offspring) using a 50K SNP panel. The GWAS were performed using all genotyped and phenotyped animals. The genotyped imputation was performed from LD panels (LD0.5K, LD1K and LD3K) to high-density panel (HD), using information from parents and 20% of offspring in the reference set and the remaining 80% in the validation set. In addition, we tested the accuracy of genomic selection using true and imputed genotypes comparing the accuracy obtained from pedigree-based best linear unbiased prediction (PBLUP) and genomic predictions. The results from GWAS supports evidence of the polygenic nature of FY and HW. The accuracy of imputation ranged from 0.90 to 0.98 for LD0.5K and LD3K, respectively. The accuracy of genomic prediction outperformed the estimated breeding value from PBLUP. The use of imputation for genomic selection resulted in an increased relative accuracy independent of the trait and LD panel analyzed. The present results suggest that genotype imputation could be a cost-effective strategy for genomic selection in Nile tilapia breeding programs.

Intermediate wheatgrass (Thinopyrum intermedium, IWG) is a perennial grain crop with high biomass and grain yield, long seeds, and resistance to pests and diseases. It also reduces soil erosion, nitrate and mineral leaching into underground water tables, and sequesters carbon in its roots. The domestication timeline of IWG as a grain crop spans only 3 decades, hence it lags annual grain crops in yield and seed characteristics. One approach to improve its agronomic traits is by using molecular markers to uncover marker-trait associations. In this study, we performed association mapping on IWG breeding germplasm from the third recurrent selection cycle at the University of Minnesota. The IWG population was phenotyped in St Paul, MN in 2017 and 2018, and in Crookston, MN in 2018 for grain yield, seed length, width and weight, spike length and weight, and number of spikelets per spike. Strong positive correlations were observed among most trait pairs, with correlations as high as 0.76. Genotyping using high throughput sequencing identified 8,899 high-quality genome-wide SNPs which were combined with phenotypic data in association mapping to discover regions associated with the yield component traits. We detected 154 genetic loci associated with these traits of which 19 were shared between at least two traits. Prediction of breeding values using significant loci as fixed effects in genomic selection model improved predictive abilities by up to 14%. Genetic mapping of agronomic traits followed by using genomic selection to predict breeding values can assist breeders in selecting superior genotypes to accelerate IWG domestication.

The genetic merit of individuals can be estimated using models with dense markers and pedigree information. Early genomic models accounted only for additive effects. However, the prediction of non-additive effects is important for different forest breeding systems where the whole genotypic value can be captured through clonal propagation. In this study, we evaluated the integration of marker data with pedigree information, in models that included or ignored non-additive effects. We tested the models Reproducing Kernel Hilbert Spaces (RKHS) and BayesA, with additive and additive-dominance frameworks. Model performance was assessed for the traits tree height, diameter at breast height and rust resistance, measured in 923 pine individuals from a structured population of 71 full-sib families. We have also simulated a population with similar genetic properties and evaluated the performance of models for six simulated traits with distinct genetic architectures. Different cross validation strategies were evaluated, and highest accuracies were achieved using within family cross validation. The inclusion of pedigree information in genomic prediction models did not yield higher accuracies. The different RKHS models resulted in similar predictions accuracies, and RKHS and BayesA generated substantially better predictions than pedigree-only models. The additive-BayesA resulted in higher accuracies than RKHS for rust incidence and in simulated additive-oligogenic traits. For DBH, HT and additive-dominance polygenic traits, the RKHS- based models showed slightly higher accuracies than BayesA. Our results indicate that BayesA performs the best for traits with few genes with major effects, while RKHS based models can best predict genotypic effects for clonal selection of complex traits.

Genomic selection is an efficient approach to get shorter breeding cycles in recurrent selection programs and greater genetic gains with selection of superior individuals. Despite advances in genotyping techniques, genetic studies for polyploid species have been limited to a rough approximation of studies in diploid species. The major challenge is to distinguish the different types of heterozygotes present in polyploid populations. In this work, we evaluated different genomic prediction models applied to a recurrent selection population of 530 genotypes of Panicum maximum, an autotetraploid forage grass. We also investigated the effect of the allele dosage in the prediction, i.e., considering tetraploid (GS-TD) or diploid (GS-DD) allele dosage. A longitudinal linear mixed model was fitted for each one of the six phenotypic traits, considering different covariance matrices for genetic and residual effects. A total of 41,424 genotyping-by-sequencing markers were obtained using 96-plex and Pst1 restriction enzyme, and quantitative genotype calling was performed. Six predictive models were generalized to tetraploid species and predictive ability was estimated by a replicated fivefold cross-validation process. GS-TD and GS-DD models were performed considering 1,223 informative markers. Overall, GS-TD data yielded higher predictive abilities than with GS-DD data. However, different predictive models had similar predictive ability performance. In this work, we provide bioinformatic and modeling guidelines to consider tetraploid dosage and observed that genomic selection may lead to additional gains in recurrent selection program of P. maximum.

The lowland ecotype of switchgrass has generated considerable interest because of its higher biomass yield and late flowering characteristics compared to the upland ecotype. However, lowland ecotypes planted in northern latitudes exhibit very low winter survival. Implementation of genomic selection could potentially enhance switchgrass breeding for winter survival by reducing generation time while eliminating the dependence on weather. The objectives of this study were to assess the potential of genomic selection for winter survival in lowland switchgrass by combining multiple populations in the training set and applying the selected model in two independent testing datasets for validation. Marker data were generated using exome capture sequencing. Validation was conducted using (1) indirect indicators of winter adaptation based on geographic and climatic variables of accessions from different source locations and (2) winter survival estimates of the phenotype. The prediction accuracies were significantly higher when the training dataset comprising all populations was used in fivefold cross validation but its application was not useful in the independent validation dataset. Nevertheless, modeling for population heterogeneity improved the prediction accuracy to some extent but the genetic relationship between the training and validation populations was found to be more influential. The predicted winter survival of lowland switchgrass indicated latitudinal and longitudinal variability, with the northeast USA the region for most cold tolerant lowland populations. Our results suggested that GS could provide valuable opportunities for improving winter survival and accelerate the lowland switchgrass breeding programs toward the development of cold tolerant cultivars suitable for northern latitudes.

Developing training sets for genomic prediction in hybrid crops requires producing hybrid seed for a large number of entries. In autogamous crop species (e.g., wheat, rice, rapeseed, cotton) this requires elaborate hybridization systems to prevent self-pollination and presents a significant impediment to the implementation of hybrid breeding in general and genomic selection in particular. An alternative to F1 hybrids are bulks of F2 seed from selfed F1 plants (F1:2). Seed production for F1:2 bulks requires no hybridization system because the number of F1 plants needed for producing enough F1:2 seed for multi-environment testing can be generated by hand-pollination. This study evaluated the suitability of F1:2 bulks for use in training sets for genomic prediction of F1 level general combining ability and hybrid performance, under different degrees of divergence between heterotic groups and modes of gene action, using quantitative genetic theory and simulation of a genomic prediction experiment. The simulation, backed by theory, showed that F1:2 training sets are expected to have a lower prediction accuracy relative to F1 training sets, particularly when heterotic groups have strongly diverged. The accuracy penalty, however, was only modest and mostly because of a lower heritability, rather than because of a difference in F1 and F1:2 genetic values. It is concluded that resorting to F1:2 bulks is, in theory at least, a promising approach to remove the significant complication of a hybridization system from the breeding process.

Multiple-trait experiments with mixed phenotypes (binary, ordinal and continuous) are not rare in animal and plant breeding programs. However, there is a lack of statistical models that can exploit the correlation between traits with mixed phenotypes in order to improve prediction accuracy in the context of genomic selection (GS). For this reason, when breeders have mixed phenotypes, they usually analyze them using univariate models, and thus are not able to exploit the correlation between traits, which many times helps improve prediction accuracy. In this paper we propose applying deep learning for analyzing multiple traits with mixed phenotype data in terms of prediction accuracy. The prediction performance of multiple-trait deep learning with mixed phenotypes (MTDLMP) models was compared to the performance of univariate deep learning (UDL) models. Both models were evaluated using predictors with and without the genotype × environment (G×E) interaction term (I and WI, respectively). The metric used for evaluating prediction accuracy was Pearson’s correlation for continuous traits and the percentage of cases correctly classified (PCCC) for binary and ordinal traits. We found that a modest gain in prediction accuracy was obtained only in the continuous trait under the MTDLMP model compared to the UDL model, whereas for the other traits (1 binary and 2 ordinal) we did not find any difference between the two models. In both models we observed that the prediction performance was better for WI than for I. The MTDLMP model is a good alternative for performing simultaneous predictions of mixed phenotypes (binary, ordinal and continuous) in the context of GS.

Evidence that genomic selection (GS) is a technology that is revolutionizing plant breeding continues to grow. However, it is very well documented that its success strongly depends on statistical models, which are used by GS to perform predictions of candidate genotypes that were not phenotyped. Because there is no universally better model for prediction and models for each type of response variable are needed (continuous, binary, ordinal, count, etc.), an active area of research aims to develop statistical models for the prediction of univariate and multivariate traits in GS. However, most of the models developed so far are for univariate and continuous (Gaussian) traits. Therefore, to overcome the lack of multivariate statistical models for genome-based prediction by improving the original version of the BMTME, we propose an improved Bayesian multi-trait and multi-environment (BMTME) R package for analyzing breeding data with multiple traits and multiple environments. We also introduce Bayesian multi-output regressor stacking (BMORS) functions that are considerably efficient in terms of computational resources. The package allows parameter estimation and evaluates the prediction performance of multi-trait and multi-environment data in a reliable, efficient and user-friendly way. We illustrate the use of the BMTME with real toy datasets to show all the facilities that the software offers the user. However, for large datasets, the BME() and BMTME() functions of the BMTME R package are very intense in terms of computing time; on the other hand, less intensive computing is required with BMORS functions BMORS() and BMORS_Env() that are also included in the BMTME package.

Hyperspectral reflectance phenotyping and genomic selection are two emerging technologies that have the potential to increase plant breeding efficiency by improving prediction accuracy for grain yield. Hyperspectral cameras quantify canopy reflectance across a wide range of wavelengths that are associated with numerous biophysical and biochemical processes in plants. Genomic selection models utilize genome-wide marker or pedigree information to predict the genetic values of breeding lines. In this study, we propose a multi-kernel GBLUP approach to genomic selection that uses genomic marker-, pedigree-, and hyperspectral reflectance-derived relationship matrices to model the genetic main effects and genotype × environment (G×E) interactions across environments within a bread wheat (Triticum aestivumL.) breeding program. We utilized an airplane equipped with a hyperspectral camera to phenotype five differentially managed treatments of the yield trials conducted by the Bread Wheat Improvement Program, International Maize and Wheat Improvement Center (CIMMYT) at Ciudad Obregón, México over four breeding cycles. We observed that single-kernel models using hyperspectral reflectance-derived relationship matrices performed similarly or superior to marker-and pedigree-based genomic selection models when predicting within and across environments. Multi-kernel models combining marker/pedigree information with hyperspectral reflectance phentoypes had the highest prediction accuracies; however, improvements in accuracy over marker-and pedigree-based models were marginal when correcting for days to heading. Our results demonstrates the potential of hyperspectral imaging in predicting grain yield within a multi-environment context, it also supports further studies on integration of hyperspectral reflectance phenotyping in breeding programs.

Two key features of recombinant inbred panels are well-characterized genomes and reproducibility. Here we report on the sequenced genomes of six additional Collaborative Cross (CC) strains and on inbreeding progress of 72 CC strains. We have previously reported on the sequences of 69 CC strains that were publicly available, bringing the total of CC strains with whole genome sequence up to 75. The sequencing of these six CC strains updates the efforts toward inbreeding undertaken by the UNC Systems Genetics Core. The timing reflects our competing mandates to release to the public as many CC strains as possible while achieving an acceptable level of inbreeding. The new six strains have a higher than average founder contribution from non-domesticus strains than the previously released CC strains. Five of the six strains also have high residual heterozygosity (>14%), which may be related to non-domesticus founder contributions. Finally, we report on updated estimates on residual heterozygosity across the entire CC population using a novel, simple and cost effective genotyping platform on three mice from each strain. We observe a reduction in residual heterozygosity across all previously released CC strains. We discuss the optimal use of different genetic resources available for the CC population.

Genomic prediction is a useful tool to accelerate genetic gain in selection using DNA marker information. However, this technology typically relies on standard prediction procedures, such as genomic BLUP, that are not designed to accommodate population heterogeneity resulting from differences in marker effects across populations. In this study, we assayed different prediction procedures to capture marker-by-population interactions in genomic prediction models. Prediction procedures included genomic BLUP and two kernel-based extensions of genomic BLUP which explicitly accounted for population heterogeneity. To model population heterogeneity, dissemblance between populations was either depicted by a unique coefficient (as previously reported), or a more flexible function of genetic distance between populations (proposed herein). Models under investigation were applied in a diverse switchgrass sample under two validation schemes: whole-sample calibration, where all individuals except selection candidates are included in the calibration set, and cross-population calibration, where the target population is entirely excluded from the calibration set. First, we showed that using fixed effects, from principal components or putative population groups, appeared detrimental to prediction accuracy, especially in cross-population calibration. Then we showed that modeling population heterogeneity by our proposed procedure resulted in highly significant improvements in model fit. In such cases, gains in accuracy were often positive. These results suggest that population heterogeneity may be parsimoniously captured by kernel methods. However, in cases where improvement in model fit by our proposed procedure is null-to-moderate, ignoring heterogeneity should probably be preferred due to the robustness and simplicity of the standard genomic BLUP model.

Estimation of allele dosage, using genomic data, in autopolyploids is challenging and current methods often result in the misclassification of genotypes. Some progress has been made when using SNP arrays, but the major challenge is when using next generation sequencing data. Here we compare the use of read depth as continuous parameterization with ploidy parameterizations in the context of genomic selection (GS). Additionally, different sources of information to build relationship matrices were compared. A real breeding population of the autotetraploid species blueberry (Vaccinium corybosum), composed of 1,847 individuals was phenotyped for eight yield and fruit quality traits over two years. Continuous genotypic based models performed as well as the best models. This approach also reduces the computational time and avoids problems associated with misclassification of genotypic classes when assigning dosage in polyploid species. This approach could be very valuable for species with higher ploidy levels or for emerging crops where ploidy is not well understood. To our knowledge, this work constitutes the first study of genomic selection in blueberry. Accuracies are encouraging for application of GS for blueberry breeding. GS could reduce the time for cultivar release by three years, increasing the genetic gain per cycle by 86% on average when compared to phenotypic selection, and 32% when compared with pedigree-based selection. Finally, the genotypic and phenotypic data used in this study are made available for comparative analysis of dosage calling and genomic selection prediction models in the context of autopolyploids.

Diverse crops are both outbred and clonally propagated. Breeders typically use truncation selection of parents and invest significant time, land and money evaluating the progeny of crosses to find exceptional genotypes. We developed and tested genomic mate selection criteria suitable for organisms of arbitrary homozygosity level where the full-sibling progeny are of direct interest as future parents and/or cultivars. We extended cross variance and covariance variance prediction to include dominance effects and predicted the multivariate selection index genetic variance of crosses based on haplotypes of proposed parents, marker effects and recombination frequencies. We combined the predicted mean and variance into usefulness criteria for parent and variety development. We present an empirical study of cassava (Manihot esculenta), a staple tropical root crop. We assessed the potential to predict the multivariate genetic distribution (means, variances and trait covariances) of 462 cassava families in terms of additive and total value using cross-validation. We were able to predict all genetic variances and most covariances with non-zero accuracy. We also tested a directional dominance model and found significant inbreeding depression for most traits and a boost in total merit accuracy for root yield. We predicted 47,083 possible crosses of 306 parents and contrasted them to those previously tested to show how mate selection can reveal new potential within the germplasm. We enable breeders to consider the potential of crosses to produce future parents (progeny with excellent breeding values) and varieties (progeny with top performance).Author SummaryBreeders typically use truncation selection and invest significant resources evaluating progeny to find exceptional genotypes. We extended genetic variance and trait covariance prediction to include dominance and predicting the multivariate selection index variance. We enable mate selection based on potential to produce future parents (progeny with excellent breeding values) and/or varieties (progeny with top performance). Using cross-validation, we demonstrate that genetic variances and covariances can be predicted with non-zero accuracy in cassava, a staple tropical root crop.

Oat (Avena sativa L.) seed is a rich resource of beneficial lipids, soluble fiber, protein, and antioxidants, and is considered a healthful food for humans. Little is known regarding the genetic controllers of variation for these compounds in oat seed. We characterized natural variation in the mature seed metabolome using untargeted metabolomics on 367 diverse lines and leveraged this information to improve prediction for seed quality traits. We used a latent factor approach to define unobserved variables that may drive covariance among metabolites. One hundred latent factors were identified, of which 21% were enriched for compounds associated with lipid metabolism. Through a combination of whole-genome regression and association mapping, we show that latent factors that generate covariance for many metabolites tend to have a complex genetic architecture. Nonetheless, we recovered significant associations for 23% of the latent factors. These associations were used to inform a multi-kernel genomic prediction model, which was used to predict seed lipid and protein traits in two independent studies. Predictions for 8 of the 12 traits were significantly improved compared to genomic best linear unbiased prediction when this prediction model was informed using associations from lipid-enriched factors. This study provides new insights into variation in the oat seed metabolome and provides genomic resources for breeders to improve selection for health-promoting seed quality traits. More broadly, we outline an approach to distill high-dimensional “omics” data to a set of biologically meaningful variables and translate inferences on these data into improved breeding decisions.

In genomic selection choosing the statistical machine learning model is of paramount importance. In this paper, we present an application of a zero altered random forest model with two versions (ZAP_RF and ZAPC_RF) to deal with excess zeros in count response variables. The proposed model was compared with the conventional random forest (RF) model and with the conventional Generalized Poisson Ridge regression (GPR) using two real datasets, and we found that, in terms of prediction performance, the proposed zero inflated random forest model outperformed the conventional RF and GPR models.

Bayesian regression methods that incorporate different mixture priors for marker effects are used in multi-trait genomic prediction. These methods can also be extended to genome-wide association studies (GWAS). In multiple-trait GWAS, incorporating the underlying causal structures among traits is essential for comprehensively understanding the relationship between genotypes and traits of interest. Therefore, we develop a GWAS methodology, SEM-BayesCΠ, which, by applying the structural equation model (SEM), can be used to incorporate causal structures into a multi-trait Bayesian regression method using mixture priors. The performance of SEM-BayesCΠ was demonstrated by comparing its GWAS results with those from multi-trait BayesCΠ. Through the inductive causation (IC) algorithm, three potential causal structures were inferred of 0.9 highest posterior density (HPD) interval. SEM-BayesCΠ provides a more comprehensive understanding of the genotype-phenotype mapping than multi-trait BayesCΠ by performing GWAS based on indirect, direct and overall marker effects. The software tool JWAS offers open-source routines to perform these analyses.

Genomic prediction uses DNA sequences and phenotypes to predict genetic values. In homogeneous populations, theory indicates that the accuracy of genomic prediction increases with sample size. However, differences in allele frequencies and linkage disequilibrium patterns can lead to heterogeneity in SNP effects. In this context, calibrating genomic predictions using a large, potentially heterogeneous, training data set may not lead to optimal prediction accuracy. Some studies tried to address this sample size/homogeneity trade-off using training set optimization algorithms; however, this approach assumes that a single training data set is optimum for all individuals in the prediction set. Here, we propose an approach that identifies, for each individual in the prediction set, a subset from the training data (i.e., a set of support points) from which predictions are derived. The methodology that we propose is a sparse selection index (SSI) that integrates selection index methodology with sparsity-inducing techniques commonly used for high-dimensional regression. The sparsity of the resulting index is controlled by a regularization parameter (λ); the G-Best Linear Unbiased Predictor (G-BLUP) (the prediction method most commonly used in plant and animal breeding) appears as a special case which happens when λ = 0. In this study, we present the methodology and demonstrate (using two wheat data sets with phenotypes collected in 10 different environments) that the SSI can achieve significant (anywhere between 5 and 10%) gains in prediction accuracy relative to the G-BLUP.

The natural genetic diversity of agricultural species is an essential genetic resource for breeding programs aiming to improve their ecosystem and production services. A large natural ecotype diversity is usually available for most grassland species. This could be used to recombine natural climatic adaptations and agronomic value to create improved populations of grassland species adapted to future regional climates. However describing natural genetic resources can be long and costly. Molecular markers may provide useful information to help this task. This opportunity was investigated for Lolium perenne L., using a set of 385 accessions from the natural diversity of this species collected right across Europe and provided by genebanks of several countries. For each of these populations, genotyping provided the allele frequencies of 189,781 SNP markers. GWAS were implemented for over 30 agronomic and/or putatively adaptive traits recorded in three climatically contrasted locations (France, Belgium, Germany). Significant associations were detected for hundreds of markers despite a strong confounding effect of the genetic background; most of them pertained to phenology traits. It is likely that genetic variability in these traits has had an important contribution to environmental adaptation and ecotype differentiation. Genomic prediction models calibrated using natural diversity were found to be highly effective to describe natural populations for almost all traits as well as commercial synthetic populations for some important traits such as disease resistance, spring growth or phenological traits. These results will certainly be valuable information to help the use of natural genetic resources of other species.

Anterior cruciate ligament (ACL) rupture is a common, debilitating condition that leads to early-onset osteoarthritis and reduced quality of human life. ACL rupture is a complex disease with both genetic and environmental risk factors. Characterizing the genetic basis of ACL rupture would provide the ability to identify individuals that have high genetic risk and allow the opportunity for preventative management. Spontaneous ACL rupture is also common in dogs and shows a similar clinical presentation and progression. Thus, the dog has emerged as an excellent genomic model for human ACL rupture. Genome-wide association studies (GWAS) in the dog have identified a number of candidate genetic variants, but research in genomic prediction has been limited. In this analysis, we explore several Bayesian and machine learning models for genomic prediction of ACL rupture in the Labrador Retriever dog. Our work demonstrates the feasibility of predicting ACL rupture from SNPs in the Labrador Retriever model with and without consideration of non-genetic risk factors. Genomic prediction including non-genetic risk factors approached clinical relevance using multiple linear Bayesian and non-linear models. This analysis represents the first steps toward development of a predictive algorithm for ACL rupture in the Labrador Retriever model. Future work may extend this algorithm to other high-risk breeds of dog. The ability to accurately predict individual dogs at high risk for ACL rupture would identify candidates for clinical trials that would benefit both veterinary and human medicine.

We investigated the effectiveness of mate allocation strategies accounting for non-additive genetic effects to improve crossbred performance in a two-way crossbreeding scheme. We did this by computer simulation of 10 generations of evaluation and selection. QTL effects were simulated as correlated across purebreds and crossbreds, and (positive) heterosis was simulated as directional dominance. The purebred-crossbred correlation was 0.30 or 0.68 depending on the genetic variance component used. Dominance and additive marker effects were estimated simultaneously for purebreds and crossbreds by multiple trait genomic BLUP. Four scenarios that differ in the sources of information (only purebred data, or purebred and crossbred data) and mate allocation strategies (mating at random, minimizing expected future inbreeding, or maximizing the expected total genetic value of crossbred animals) were evaluated under different cases of genetic variance components. Selecting purebred animals for purebred performance yielded a response of 0.2 genetic standard deviations of the trait “crossbred performance” per generation, whereas selecting purebred animals for crossbred performance doubled the genetic response. Mate allocation strategy to maximize the expected total genetic value of crossbred descendants resulted in a slight increase (0.8%, 4% and 0.5% depending on the genetic variance components) of the crossbred performance. Purebred populations increased homozygosity, but the heterozygosity of the crossbreds remained constant. When purebred-crossbred genetic correlation is low, selecting purebred animals for crossbred performance using crossbred information is a more efficient strategy to exploit heterosis and increase performance at the crossbred commercial level, whereas mate allocation did not improve crossbred performance.

Polygenic risk scores (PRS) use the results of genome-wide association studies (GWAS) to predict quantitative phenotypes or disease risk at an individual level, and provide a potential route to the use of genetic data in personalized medical care. However, a major barrier to the use of PRS is that the majority of GWAS come from cohorts of European ancestry. The predictive power of PRS constructed from these studies is substantially lower in non-European ancestry cohorts, although the reasons for this are unclear. To address this question, we investigate the performance of PRS for height in cohorts with admixed African and European ancestry, allowing us to evaluate ancestry-related differences in PRS predictive accuracy while controlling for environment and cohort differences. We first show that the predictive accuracy of height PRS increases linearly with European ancestry and is partially explained by European ancestry segments of the admixed genomes. We show that recombination rate, differences in allele frequencies, and differences in marginal effect sizes across ancestries all contribute to the decrease in predictive power, but none of these effects explain the decrease on its own. Finally, we demonstrate that prediction for admixed individuals can be improved by using a linear combination of PRS that includes ancestry-specific effect sizes, although this approach is at present limited by the small size of non-European ancestry discovery cohorts.

Genomic selection has been successfully implemented in plant and animal breeding. The transition of parental selection based on phenotypic characteristics to genomic selection (GS) has reduced breeding time and cost while accelerating the rate of genetic progression. Although breeding methods have been adapted to include genomic selection, parental selection often involves truncation selection, selecting the individuals with the highest genomic estimated breeding values (GEBVs) in the hope that favorable properties will be passed to their offspring. This ensures genetic progression and delivers offspring with high genetic values. However, several favorable quantitative trait loci (QTL) alleles risk being eliminated from the breeding population during breeding. We show that this could reduce the mean genetic value that the breeding population could reach in the long term with up to 40%. In this paper, by means of a simulation study, we propose a new method for parental mating that is able to preserve the genetic variation in the breeding population, preventing premature convergence of the genetic values to a local optimum, thus maximizing the genetic values in the long term. We do not only prevent the fixation of several unfavorable QTL alleles, but also demonstrate that the genetic values can be increased by up to 15 percentage points compared with truncation selection.

Genomic selection uses whole-genome marker models to predict phenotypes or genetic values for complex traits. Some of these models fit interaction terms between markers, and are therefore called epistatic. The biological interpretation of the corresponding fitted effects is not straightforward and there is the threat of overinterpreting their functional meaning. Here we show that the predictive ability of epistatic models relative to additive models can change with the density of the marker panel. In more detail, we show that for publicly available Arabidopsis and rice datasets, an initial superiority of epistatic models over additive models, which can be observed at a lower marker density, vanishes when the number of markers increases. We relate these observations to earlier results reported in the context of association studies which showed that detecting statistical epistatic effects may not only be related to interactions in the underlying genetic architecture, but also to incomplete linkage disequilibrium at low marker density (“Phantom Epistasis”). Finally, we illustrate in a simulation study that due to phantom epistasis, epistatic models may also predict the genetic value of an underlying purely additive genetic architecture better than additive models, when the marker density is low. Our observations can encourage the use of genomic epistatic models with low density panels, and discourage their biological over-interpretation. Fil: Schrauf, Matías Florián. Universidad de Buenos Aires. Facultad de Agronomía. Departamento de Producción Animal. Cátedra de Mejoramiento Genético Animal; Argentina. Consejo Nacional de Investigaciones Científicas y Técnicas; Argentina Fil: Martini, Johannes W.R.. Centro Internacional de Mejoramiento de Maíz y Trigo; México Fil: Simianer, Henner. Universität Göttingen; Alemania Fil: de los Campos, Gustavo. Michigan State University; Estados Unidos Fil: Cantet, Rodolfo Juan Carlos. Consejo Nacional de Investigaciones Científicas y Técnicas. Oficina de Coordinación Administrativa Parque Centenario. Unidad Ejecutora de Investigaciones en Producción Animal. Universidad de Buenos Aires. Facultad de Ciencias Veterinarias. Unidad Ejecutora de Investigaciones en Producción Animal; Argentina Fil: Freudenthal, Jan. Universität Würzburg; Alemania Fil: Korte, Arthur. Universität Würzburg; Alemania Fil: Munilla Leguizamon, Sebastian. Universidad de Buenos Aires. Facultad de Agronomía. Departamento de Producción Animal. Cátedra de Mejoramiento Genético Animal; Argentina. Consejo Nacional de Investigaciones Científicas y Técnicas; Argentina

Weedy rice (Oryza spp.) is a problematic weed of cultivated rice (O. sativa) around the world. Recent studies have established multiple independent evolutionary origins of weedy rice, raising questions about the traits and genes that are essential for the evolution of this weed. Among world regions, South Asia stands out due to the heterogeneity of its weedy rice populations, which can be traced to at least three origins: two through de-domestication from distinct cultivated rice varieties, and one from local wild rice (O. rufipogon/O. nivara). Here we examine five traits considered typical of or advantageous to weedy rice in weedy, cultivated and wild rice samples from South Asia. We establish that convergence among all three weed groups occurs for easy seed shattering, red pericarp color, and compact plant architecture, suggesting that these traits are essential for weed success in the South Asian agricultural environment. A high degree of convergence for black hull color is also seen among weeds with wild ancestors and weeds evolved from the aus cultivated rice group. We also examine polymorphism in five known domestication candidate genes, and find that Rc and Bh4 are associated with weed seed pericarp color and hull color, respectively, and weedy alleles segregate in the ancestral populations, as do alleles for the seed dormancy-linked gene Sdr4. The presence of a domestication related allele at the seed shattering locus, sh4, in weedy rice populations with cultivated ancestry supports a de-domestication origin for these weedy groups, and raises questions about the reacquisition of the shattering trait in these weedy populations. Our characterization of weedy rice phenotypes in South Asia and their associated candidate genes contribute to the emerging understanding of the mechanisms by which weedy rice evolves worldwide, suggesting that standing ancestral variation is often the source of weedy traits in independently evolved groups, and highlighting the reservoir of genetic variation that is present in cultivated varieties as well as in wild rice, and its potential for phenotypic evolution.

Plant and animal breeders are interested in selecting the best individuals from a candidate set for the next breeding cycle. In this paper, we propose a formal method under the Bayesian decision theory framework to tackle the selection problem based on genomic selection (GS) in single- and multi-trait settings. We proposed and tested three univariate loss functions (Kullback-Leibler, KL; Continuous Ranked Probability Score, CRPS; Linear-Linear loss, LinLin) and their corresponding multivariate generalizations (Kullback-Leibler, KL; Energy Score, EnergyS; and the Multivariate Asymmetric Loss Function, MALF). We derived and expressed all the loss functions in terms of heritability and tested them on a real wheat dataset for one cycle of selection and in a simulated selection program. The performance of each univariate loss function was compared with the standard method of selection (Std) that does not use loss functions. We compared the performance in terms of the selection response and the decrease in the population’s genetic variance during recurrent breeding cycles. Results suggest that it is possible to obtain better performance in a long-term breeding program using the single-trait scheme by selecting 30% of the best individuals in each cycle but not by selecting 10% of the best individuals. For the multi-trait approach, results show that the population mean for all traits under consideration had positive gains, even though two of the traits were negatively correlated. The corresponding population variances were not statistically different from the different loss function during the 10th selection cycle. Using the loss function should be a useful criterion when selecting the candidates for selection for the next breeding cycle.

One of the major issues in plant breeding is the occurrence of genotype × environment (GE) interaction. Several models have been created to understand this phenomenon and explore it. In the genomic era, several models were employed to improve selection by using markers and account for GE interaction simultaneously. Some of these models use special genetic covariance matrices. In addition, the scale of multi-environment trials is getting larger, and this increases the computational challenges. In this context, we propose an R package that, in general, allows building GE genomic covariance matrices and fitting linear mixed models, in particular, to a few genomic GE models. Here we propose two functions: one to prepare the genomic kernels accounting for the genomic GE and another to perform genomic prediction using a Bayesian linear mixed model. A specific treatment is given for sparse covariance matrices, in particular, to block diagonal matrices that are present in some GE models in order to decrease the computational demand. In empirical comparisons with Bayesian Genomic Linear Regression (BGLR), accuracies and the mean squared error were similar; however, the computational time was up to five times lower than when using the classic approach. Bayesian Genomic Genotype × Environment Interaction (BGGE) is a fast, efficient option for creating genomic GE kernels and making genomic predictions.

Genomic selection applied to plant breeding enables earlier estimates of a line’s performance and significant reductions in generation interval. Several factors affecting prediction accuracy should be well understood if breeders are to harness genomic selection to its full potential. We used a panel of 10,375 bread wheat (Triticum aestivum) lines genotyped with 18,101 SNP markers to investigate the effect and interaction of training set size, population structure and marker density on genomic prediction accuracy. Through assessing the effect of training set size we showed the rate at which prediction accuracy increases is slower beyond approximately 2,000 lines. The structure of the panel was assessed via principal component analysis and K-means clustering, and its effect on prediction accuracy was examined through a novel cross-validation analysis according to the K-means clusters and breeding cohorts. Here we showed that accuracy can be improved by increasing the diversity within the training set, particularly when relatedness between training and validation sets is low. The breeding cohort analysis revealed that traits with higher selection pressure (lower allelic diversity) can be more accurately predicted by including several previous cohorts in the training set. The effect of marker density and its interaction with population structure was assessed for marker subsets containing between 100 and 17,181 markers. This analysis showed that response to increased marker density is largest when using a diverse training set to predict between poorly related material. These findings represent a significant resource for plant breeders and contribute to the collective knowledge on the optimal structure of calibration panels for genomic prediction.

Genomic selection enables cumulative genetic gains in key production traits such as disease resistance, playing an important role in the economic and environmental sustainability of aquaculture production. However, it requires genome-wide genetic marker data on large populations, which can be prohibitively expensive. Genotype imputation is a cost-effective method for obtaining high-density genotypes, but its value in aquaculture breeding programs which are characterized by large full-sibling families has yet to be fully assessed. The aim of this study was to optimize the use of low-density genotypes and evaluate genotype imputation strategies for cost-effective genomic prediction. Phenotypes and genotypes (78,362 SNPs) were obtained for 610 individuals from a Scottish Atlantic salmon breeding program population (Landcatch, UK) challenged with sea lice, Lepeophtheirus salmonis. The genomic prediction accuracy of genomic selection was calculated using GBLUP approaches and compared across SNP panels of varying densities and composition, with and without imputation. Imputation was tested when parents were genotyped for the optimal SNP panel, and offspring were genotyped for a range of lower density imputation panels. Reducing SNP density had little impact on prediction accuracy until 5,000 SNPs, below which the accuracy dropped. Imputation accuracy increased with increasing imputation panel density. Genomic prediction accuracy when offspring were genotyped for just 200 SNPs, and parents for 5,000 SNPs, was 0.53. This accuracy was similar to the full high density and optimal density dataset, and markedly higher than using 200 SNPs without imputation. These results suggest that imputation from very low to medium density can be a cost-effective tool for genomic selection in Atlantic salmon breeding programs.