Evaluating Methods of Updating Training Data in Long-Term Genomewide Selection

Genomic selection in plant breeding: Key factors shaping two decades of progress

Genomic selection, the application of genomic prediction (GP) models to select candidate individuals, has significantly advanced in the past two decades, effectively accelerating genetic gains in plant breeding. This article provides a holistic overview of key factors that have influenced GP in plant breeding during this period. We delved into the pivotal roles of training population size and genetic diversity, and their relationship with the breeding population, in determining GP accuracy. Special emphasis was placed on optimizing training population size. We explored its benefits and the associated diminishing returns beyond an optimum size. This was done while considering the balance between resource allocation and maximizing prediction accuracy through current optimization algorithms. The density and distribution of single-nucleotide polymorphisms, level of linkage disequilibrium, genetic complexity, trait heritability, statistical machine-learning methods, and non-additive effects are the other vital factors. Using wheat, maize, and potato as examples, we summarize the effect of these factors on the accuracy of GP for various traits. The search for high accuracy in GP-theoretically reaching one when using the Pearson's correlation as a metric-is an active research area as yet far from optimal for various traits. We hypothesize that with ultra-high sizes of genotypic and phenotypic datasets, effective training population optimization methods and support from other omics approaches (transcriptomics, metabolomics and proteomics) coupled with deep-learning algorithms could overcome the boundaries of current limitations to achieve the highest possible prediction accuracy, making genomic selection an effective tool in plant breeding.

Maximizing efficiency in sunflower breeding through historical data optimization

Genomic selection (GS) has become an increasingly popular tool in plant breeding programs, propelled by declining genotyping costs, an increase in computational power, and rediscovery of the best linear unbiased prediction methodology over the past two decades. This development has led to an accumulation of extensive historical datasets with genotypic and phenotypic information, triggering the question of how to best utilize these datasets. Here, we investigate whether all available data or a subset should be used to calibrate GS models for across-year predictions in a 7-year dataset of a commercial hybrid sunflower breeding program. We employed a multi-objective optimization approach to determine the ideal years to include in the training set (TRS). Next, for a given combination of TRS years, we further optimized the TRS size and its genetic composition. We developed the Min_GRM size optimization method which consistently found the optimal TRS size, reducing dimensionality by 20% with an approximately 1% loss in predictive ability. Additionally, the Tails_GEGVs algorithm displayed potential, outperforming the use of all data by using just 60% of it for grain yield, a high-complexity, low-heritability trait. Moreover, maximizing the genetic diversity of the TRS resulted in a consistent predictive ability across the entire range of genotypic values in the test set. Interestingly, the Tails_GEGVs algorithm, due to its ability to leverage heterogeneity, enhanced predictive performance for key hybrids with extreme genotypic values. Our study provides new insights into the optimal utilization of historical data in plant breeding programs, resulting in improved GS model predictive ability.

Portability of genomic predictions trained on sparse factorial designs across two maize silage breeding cycles

KEY MESSAGE

We validated the efficiency of genomic predictions calibrated on sparse factorial training sets to predict the next generation of hybrids and tested different strategies for updating predictions along generations. Genomic selection offers new prospects for revisiting hybrid breeding schemes by replacing extensive phenotyping of individuals with genomic predictions. Finding the ideal design for training genomic prediction models is still an open question. Previous studies have shown promising predictive abilities using sparse factorial instead of tester-based training sets to predict single-cross hybrids from the same generation. This study aims to further investigate the use of factorials and their optimization to predict line general combining abilities (GCAs) and hybrid values across breeding cycles. It relies on two breeding cycles of a maize reciprocal genomic selection scheme involving multiparental connected reciprocal populations from flint and dent complementary heterotic groups selected for silage performances. Selection based on genomic predictions trained on a factorial design resulted in a significant genetic gain for dry matter yield in the new generation. Results confirmed the efficiency of sparse factorial training sets to predict candidate line GCAs and hybrid values across breeding cycles. Compared to a previous study based on the first generation, the advantage of factorial over tester training sets appeared lower across generations. Updating factorial training sets by adding single-cross hybrids between selected lines from the previous generation or a random subset of hybrids from the new generation both improved predictive abilities. The CDmean criterion helped determine the set of single-crosses to phenotype to update the training set efficiently. Our results validated the efficiency of sparse factorial designs for calibrating hybrid genomic prediction experimentally and showed the benefit of updating it along generations.

Oracle selection provides insight into how far off practice is from Utopia in plant breeding

Since the introduction of genomic selection in plant breeding, high genetic gains have been realized in different plant breeding programs. Various methods based on genomic estimated breeding values (GEBVs) for selecting parental lines that maximize the genetic gain as well as methods for improving the predictive performance of genomic selection have been proposed. Unfortunately, it remains difficult to measure to what extent these methods really maximize long-term genetic values. In this study, we propose oracle selection, a hypothetical frame of mind that uses the ground truth to optimally select parents or optimize the training population in order to maximize the genetic gain in each breeding cycle. Clearly, oracle selection cannot be applied in a true breeding program, but allows for the assessment of existing parental selection and training population update methods and the evaluation of how far these methods are from the optimal utopian solution.

A comparison of methods for training population optimization in genomic selection

Maximizing CDmean and Avg_GRM_self were the best criteria for training set optimization. A training set size of 50-55% (targeted) or 65-85% (untargeted) is needed to obtain 95% of the accuracy.  With the advent of genomic selection (GS) as a widespread breeding tool, mechanisms to efficiently design an optimal training set for GS models became more relevant, since they allow maximizing the accuracy while minimizing the phenotyping costs. The literature described many training set optimization methods, but there is a lack of a comprehensive comparison among them. This work aimed to provide an extensive benchmark among optimization methods and optimal training set size by testing a wide range of them in seven datasets, six different species, different genetic architectures, population structure, heritabilities, and with several GS models to provide some guidelines about their application in breeding programs. Our results showed that targeted optimization (uses information from the test set) performed better than untargeted (does not use test set data), especially when heritability was low. The mean coefficient of determination was the best targeted method, although it was computationally intensive. Minimizing the average relationship within the training set was the best strategy for untargeted optimization. Regarding the optimal training set size, maximum accuracy was obtained when the training set was the entire candidate set. Nevertheless, a 50-55% of the candidate set was enough to reach 95-100% of the maximum accuracy in the targeted scenario, while we needed a 65-85% for untargeted optimization. Our results also suggested that a diverse training set makes GS robust against population structure, while including clustering information was less effective. The choice of the GS model did not have a significant influence on the prediction accuracies.

Genomic prediction of switchgrass winter survivorship across diverse lowland populations

In the North-Central United States, lowland ecotype switchgrass can increase yield by up to 50% compared with locally adapted but early flowering cultivars. However, lowland ecotypes are not winter tolerant. The mechanism for winter damage is unknown but previously has been associated with late flowering time. This study investigated heading date (measured for two years) and winter survivorship (measured for three years) in a multi-generation population generated from two winter-hardy lowland individuals and diverse southern lowland populations. Sequencing data (311,776 markers) from 1,306 individuals were used to evaluate genome-wide trait prediction through cross-validation and progeny prediction (n = 52). Genetic variance for heading date and winter survivorship was additive with high narrow-sense heritability (0.64 and 0.71, respectively) and reliability (0.68 and 0.76, respectively). The initial negative correlation between winter survivorship and heading date degraded across generations (F1r = -0.43, pseudo-F2r = -0.28, pseudo-F2 progeny r = -0.15). Within-family predictive ability was moderately high for heading date and winter survivorship (0.53 and 0.52, respectively). A multi-trait model did not improve predictive ability for either trait. Progeny predictive ability was 0.71 for winter survivorship and 0.53 for heading date. These results suggest that lowland ecotype populations can obtain sufficient survival rates in the northern United States with two or three cycles of effective selection. Despite accurate genomic prediction, naturally occurring winter mortality successfully isolated winter tolerant genotypes and appears to be an efficient method to develop high-yielding, cold-tolerant switchgrass cultivars.

Genomic selection strategies to increase genetic gain in tea breeding programs

Tea [Camellia sinensis (L.) O. Kuntze] is mainly grown in low- to middle-income countries (LMIC) and is a global commodity. Breeding programs in these countries face the challenge of increasing genetic gain because the accuracy of selecting superior genotypes is low and resources are limited. Phenotypic selection (PS) is traditionally the primary method of developing improved tea varieties and can take over 16 yr. Genomic selection (GS) can be used to improve the efficiency of tea breeding by increasing selection accuracy and shortening the generation interval and breeding cycle. Our main objective was to investigate the potential of implementing GS in tea-breeding programs to speed up genetic progress despite the low cost of PS in LMIC. We used stochastic simulations to compare three GS-breeding programs with a Pedigree and PS program. The PS program mimicked a practical commercial tea-breeding program over a 40-yr breeding period. All the GS programs achieved at least 1.65 times higher genetic gains than the PS program and 1.4 times compared with Seed-Ped program. Seed-GSc was the most cost-effective strategy of implementing GS in tea-breeding programs. It introduces GS at the seedlings stage to increase selection accuracy early in the program and reduced the generation interval to 2 yr. The Seed-Ped program outperformed PS by 1.2 times and could be implemented where it is not possible to use GS. Our results indicate that GS could be used to improve genetic gain per unit time and cost even in cost-constrained tea-breeding programs.

Genotyping marker density and prediction models effects in long-term breeding schemes of cross-pollinated crops

In genomic recurrent selection, the more markers, the better because they buffer the linkage disequilibrium losses caused by recombination over cycles, and consequently, provide higher responses to selection. Reductions of genotyping marker density have been extensively evaluated as potential strategies to reduce the genotyping costs of genomic selection (GS). Low-density marker panels are appealing in GS because they entail lower multicollinearity and computing time and allow more individuals to be genotyped for the same cost. However, statistical models used in GS are usually evaluated with empirical data, using "static" training sets and populations. This may be adequate for making predictions during a breeding program's initial cycles but not for the long-term. Moreover, studies that focus on long selective breeding cycles generally do not consider GS models with the effect of dominance, which is particularly important for breeding outcomes in cross-pollinated crops. Hence, dominance effects are important and unexplored in GS for long-term programs involving allogamous species. To address it, we employed two approaches: analysis of empirical maize datasets and simulations of long-term breeding applying phenotypic and genomic recurrent selection (intrapopulation and reciprocal schemes). In both schemes, we simulated twenty breeding cycles and assessed the effect of marker density reduction on the population mean, the best crosses, additive variance, selective accuracy, and response to selection with models [additive, additive-dominant, general (GCA), and this plus specific combining ability (GCA + SCA)]. Our results indicate that marker reduction based on linkage disequilibrium levels provides useful predictions only within a cycle, as accuracy significantly decreases over cycles. In the long-term, without training set updating, high-marker density provides the best responses to selection. The model to be used depends on the breeding scheme: additive for intrapopulation and additive-dominant or GCA + SCA for reciprocal.

Optimizing self-pollinated crop breeding employing genomic selection: From schemes to updating training sets

Long-term breeding schemes using genomic selection (GS) can boost the response to selection per year. Although several studies have shown that GS delivers a higher response to selection, only a few analyze which stage GS produces better results and how to update the training population to maintain prediction accuracy. We used stochastic simulation to compare five GS breeding schemes in a self-pollinated long-term breeding program. Also, we evaluated four strategies, using distinct methods and sizes, to update the training set. Finally, regarding breeding schemes, we proposed a new approach using GS to select the best individuals in each F2 progeny, based on genomic estimated breeding values and genetic divergence, to cross them and generate a new recombination event. Our results showed that the best scenario was using GS in F2, followed by the phenotypic selection of new parents in F4. For TS updating, adding new data every cycle (over 768) to update the TS maintains the prediction accuracy at satisfactory levels for more breeding cycles. However, only the last three generations can be kept in the TS, optimizing the genetic relationship between TS and the targeted population and reducing the computing demand and risks. Hence, we believe that our results may help breeders optimize GS in their programs and improve genetic gain in long-term schemes.

Building a Calibration Set for Genomic Prediction, Characteristics to Be Considered, and Optimization Approaches

The efficiency of genomic selection strongly depends on the prediction accuracy of the genetic merit of candidates. Numerous papers have shown that the composition of the calibration set is a key contributor to prediction accuracy. A poorly defined calibration set can result in low accuracies, whereas an optimized one can considerably increase accuracy compared to random sampling, for a same size. Alternatively, optimizing the calibration set can be a way of decreasing the costs of phenotyping by enabling similar levels of accuracy compared to random sampling but with fewer phenotypic units. We present here the different factors that have to be considered when designing a calibration set, and review the different criteria proposed in the literature. We classified these criteria into two groups: model-free criteria based on relatedness, and criteria derived from the linear mixed model. We introduce criteria targeting specific prediction objectives including the prediction of highly diverse panels, biparental families, or hybrids. We also review different ways of updating the calibration set, and different procedures for optimizing phenotyping experimental designs.

Deep scoping: a breeding strategy to preserve, reintroduce and exploit genetic variation

The deep scoping method incorporates the use of a gene bank together with different population layers to reintroduce genetic variation into the breeding population, thus maximizing the long-term genetic gain without reducing the short-term genetic gain or increasing the total financial cost. Genomic prediction is often combined with truncation selection to identify superior parental individuals that can pass on favorable quantitative trait locus (QTL) alleles to their offspring. However, truncation selection reduces genetic variation within the breeding population, causing a premature convergence to a sub-optimal genetic value. In order to also increase genetic gain in the long term, different methods have been proposed that better preserve genetic variation. However, when the genetic variation of the breeding population has already been reduced as a result of prior intensive selection, even those methods will not be able to avert such premature convergence. Pre-breeding provides a solution for this problem by reintroducing genetic variation into the breeding population. Unfortunately, as pre-breeding often relies on a separate breeding population to increase the genetic value of wild specimens before introducing them in the elite population, it comes with an increased financial cost. In this paper, on the basis of a simulation study, we propose a new method that reintroduces genetic variation in the breeding population on a continuous basis without the need for a separate pre-breeding program or a larger population size. This way, we are able to introduce favorable QTL alleles into an elite population and maximize the genetic gain in the short as well as in the long term without increasing the financial cost.

Subsampling and DNA pooling can increase gains through genomic selection in switchgrass

Genomic selection (GS) can accelerate breeding cycles in perennial crops such as the bioenergy grass switchgrass (Panicum virgatum L.). The sequencing costs of GS can be reduced by pooling DNA samples in the training population (TP), only sequencing TP phenotypic outliers, or pooling candidate population (CP) samples. These strategies were simulated for two traits (spring vigor and anthesis date) in three breeding populations. Sequencing only the outlier 50% of the TP phenotype distribution resulted in a penalty of <5% of the predictive ability, measured using cross-validation. Predictive ability also decreased when sequencing progressively fewer TP DNA pools, but TPs constructed from only two phenotypically contrasting DNA samples retained a mean of >80% predictive ability relative to individual TP sequencing. Novel group testing methods allowed greater than one CP individual to be screened per sequenced DNA sample but resulted in a predictive ability penalty. To determine the impact of reduced sequencing, genetic gain was calculated for seven GS scenarios with variable sequencing budgets. Reduced TP sequencing and most CP pooling methods were superior to individual sequence-based GS when sequencing resources were restricted (2,000 DNA samples per 5-yr cycle). Only one scenario was superior to individual sequencing when sequencing budgets were large (8,000 DNA samples per 5-yr cycle). This study highlights multiple routes for reduced sequencing costs in GS.

Predicting Fusarium Head Blight Resistance for Advanced Trials in a Soft Red Winter Wheat Breeding Program With Genomic Selection

Many studies have evaluated the effectiveness of genomic selection (GS) using cross-validation within training populations; however, few have looked at its performance for forward prediction within a breeding program. The objectives for this study were to compare the performance of naïve GS (NGS) models without covariates and multi-trait GS (MTGS) models by predicting two years of F_{4: 7} advanced breeding lines for three Fusarium head blight (FHB) resistance traits, deoxynivalenol (DON) accumulation, Fusarium damaged kernels (FDK), and severity (SEV) in soft red winter wheat and comparing predictions with phenotypic performance over two years of selection based on selection accuracy and response to selection. On average, for DON, the NGS model correctly selected 69.2% of elite genotypes, while the MTGS model correctly selected 70.1% of elite genotypes compared with 33.0% based on phenotypic selection from the advanced generation. During the 2018 breeding cycle, GS models had the greatest response to selection for DON, FDK, and SEV compared with phenotypic selection. The MTGS model performed better than NGS during the 2019 breeding cycle for all three traits, whereas NGS outperformed MTGS during the 2018 breeding cycle for all traits except for SEV. Overall, GS models were comparable, if not better than phenotypic selection for FHB resistance traits. This is particularly helpful when adverse environmental conditions prohibit accurate phenotyping. This study also shows that MTGS models can be effective for forward prediction when there are strong correlations between traits of interest and covariates in both training and validation populations.

Training Set Optimization for Sparse Phenotyping in Genomic Selection: A Conceptual Overview

Genomic selection (GS) is becoming an essential tool in breeding programs due to its role in increasing genetic gain per unit time. The design of the training set (TRS) in GS is one of the key steps in the implementation of GS in plant and animal breeding programs mainly because (i) TRS optimization is critical for the efficiency and effectiveness of GS, (ii) breeders test genotypes in multi-year and multi-location trials to select the best-performing ones. In this framework, TRS optimization can help to decrease the number of genotypes to be tested and, therefore, reduce phenotyping cost and time, and (iii) we can obtain better prediction accuracies from optimally selected TRS than an arbitrary TRS. Here, we concentrate the efforts on reviewing the lessons learned from TRS optimization studies and their impact on crop breeding and discuss important features for the success of TRS optimization under different scenarios. In this article, we review the lessons learned from training population optimization in plants and the major challenges associated with the optimization of GS including population size, the relationship between training and test set (TS), update of TRS, and the use of different packages and algorithms for TRS implementation in GS. Finally, we describe general guidelines to improving the rate of genetic improvement by maximizing the use of the TRS optimization in the GS framework.

TrainSel: An R Package for Selection of Training Populations

A major barrier to the wider use of supervised learning in emerging applications, such as genomic selection, is the lack of sufficient and representative labeled data to train prediction models. The amount and quality of labeled training data in many applications is usually limited and therefore careful selection of the training examples to be labeled can be useful for improving the accuracies in predictive learning tasks. In this paper, we present an R package, TrainSel, which provides flexible, efficient, and easy-to-use tools that can be used for the selection of training populations (STP). We illustrate its use, performance, and potentials in four different supervised learning applications within and outside of the plant breeding area.

Heterosis and Hybrid Crop Breeding: A Multidisciplinary Review

Although hybrid crop varieties are among the most popular agricultural innovations, the rationale for hybrid crop breeding is sometimes misunderstood. Hybrid breeding is slower and more resource-intensive than inbred breeding, but it allows systematic improvement of a population by recurrent selection and exploitation of heterosis simultaneously. Inbred parental lines can identically reproduce both themselves and their F1 progeny indefinitely, whereas outbred lines cannot, so uniform outbred lines must be bred indirectly through their inbred parents to harness heterosis. Heterosis is an expected consequence of whole-genome non-additive effects at the population level over evolutionary time. Understanding heterosis from the perspective of molecular genetic mechanisms alone may be elusive, because heterosis is likely an emergent property of populations. Hybrid breeding is a process of recurrent population improvement to maximize hybrid performance. Hybrid breeding is not maximization of heterosis per se, nor testing random combinations of individuals to find an exceptional hybrid, nor using heterosis in place of population improvement. Though there are methods to harness heterosis other than hybrid breeding, such as use of open-pollinated varieties or clonal propagation, they are not currently suitable for all crops or production environments. The use of genomic selection can decrease cycle time and costs in hybrid breeding, particularly by rapidly establishing heterotic pools, reducing testcrossing, and limiting the loss of genetic variance. Open questions in optimal use of genomic selection in hybrid crop breeding programs remain, such as how to choose founders of heterotic pools, the importance of dominance effects in genomic prediction, the necessary frequency of updating the training set with phenotypic information, and how to maintain genetic variance and prevent fixation of deleterious alleles.

Improving Genomic Prediction of Crossbred and Purebred Dairy Cattle

This study assessed the accuracy and bias of genomic prediction (GP) in purebred Holstein (H) and Jersey (J) as well as crossbred (H and J) validation cows using different reference sets and prediction strategies. The reference sets were made up of different combinations of 36,695 H and J purebreds and crossbreds. Additionally, the effect of using different sets of marker genotypes on GP was studied (conventional panel: 50k, custom panel enriched with, or close to, causal mutations: XT_50k, and conventional high-density with a limited custom set: pruned HDnGBS). We also compared the use of genomic best linear unbiased prediction (GBLUP) and Bayesian (emBayesR) models, and the traits tested were milk, fat, and protein yields. On average, by including crossbred cows in the reference population, the prediction accuracies increased by 0.01-0.08 and were less biased (regression coefficient closer to 1 by 0.02-0.16), and the benefit was greater for crossbreds compared to purebreds. The accuracy of prediction increased by 0.02 using XT_50k compared to 50k genotypes without affecting the bias. Although using pruned HDnGBS instead of 50k also increased the prediction accuracy by about 0.02, it increased the bias for purebred predictions in emBayesR models. Generally, emBayesR outperformed GBLUP for prediction accuracy when using 50k or pruned HDnGBS genotypes, but the benefits diminished with XT_50k genotypes. Crossbred predictions derived from a joint pure H and J reference were similar in accuracy to crossbred predictions derived from the two separate purebred reference sets and combined proportional to breed composition. However, the latter approach was less biased by 0.13. Most interestingly, using an equalized breed reference instead of an H-dominated reference, on average, reduced the bias of prediction by 0.16-0.19 and increased the accuracy by 0.04 for crossbred and J cows, with a little change in the H accuracy. In conclusion, we observed improved genomic predictions for both crossbreds and purebreds by equalizing breed contributions in a mixed breed reference that included crossbred cows. Furthermore, we demonstrate, that compared to the conventional 50k or high-density panels, our customized set of 50k sequence markers improved or matched the prediction accuracy and reduced bias with both GBLUP and Bayesian models.

Preservation of Genetic Variation in a Breeding Population for Long-Term Genetic Gain

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.

Effectiveness of Genomic Selection by Response to Selection for Winter Wheat Variety Improvement

Prediction performance for winter wheat grain yield and end-use quality traits. Prediction accuracies evaluated by cross-validations are significantly overestimated. Nonparametric algorithms outperform the parametric alternatives in cross-year predictions. Strategically designing training population improves response to selection. Response to selection varies across growing seasons and environments. Considering the practicality of applying genomic selection (GS) in the line development stage of a hard red winter (HRW) wheat (Triticum aestivum L.) variety development program (VDP), the effectiveness of GS was evaluated by prediction accuracy and by the response to selection across field seasons that demonstrated challenges for crop improvement under significant climate variability. Important breeding targets for wheat improvement in the southern Great Plains of the United States, including grain yield, kernel weight, wheat protein content, and sodium dodecyl sulfate (SDS) sedimentation volume as a rapid test for predicting bread-making quality, were used to estimate the effectiveness of GS across harvest years from 2014 (drought) to 2016 (normal). In general, nonparametric algorithms reproducing kernel Hilbert space (RKHS) and random forest (RF) produced higher accuracies in both same-year cross-validations (CVs) and cross-year predictions for the purpose of line selection. Further, the stability of GS performance was greatest for SDS sedimentation volume but least for wheat protein content. To ensure long-term genetic gain, our study on selection response suggested that across this sample of environmental variability, and though there are cases where phenotypic selection (PS) might be still preferred, training conducted under drought or in suboptimal conditions could provide an encouraging prediction outcome when selection decisions were made in normal conditions. However, it is not advisable to use training information collected from a normal season to predict trait performance under drought conditions. Finally, the superiority of response to selection was most evident if the training population (TP) can be optimized.

The effects of training population design on genomic prediction accuracy in wheat

Genomic selection offers several routes for increasing the genetic gain or efficiency of plant breeding programmes. In various species of livestock, there is empirical evidence of increased rates of genetic gain from the use of genomic selection to target different aspects of the breeder's equation. Accurate predictions of genomic breeding value are central to this, and the design of training sets is in turn central to achieving sufficient levels of accuracy. In summary, small numbers of close relatives and very large numbers of distant relatives are expected to enable predictions with higher accuracy. To quantify the effect of some of the properties of training sets on the accuracy of genomic selection in crops, we performed an extensive field-based winter wheat trial. In summary, this trial involved the construction of 44 F_2:4 bi- and tri-parental populations, from which 2992 lines were grown on four field locations and yield was measured. For each line, genotype data were generated for 25 K segregating SNP markers. The overall heritability of yield was estimated to 0.65, and estimates within individual families ranged between 0.10 and 0.85. Genomic prediction accuracies of yield BLUEs were 0.125-0.127 using two different cross-validation approaches and generally increased with training set size. Using related crosses in training and validation sets generally resulted in higher prediction accuracies than using unrelated crosses. The results of this study emphasise the importance of the training panel design in relation to the genetic material to which the resulting prediction model is to be applied.

Optimizing Genomic Selection for a Sorghum Breeding Program in Haiti: A Simulation Study

Young breeding programs in developing countries, like the Chibas sorghum breeding program in Haiti, face the challenge of increasing genetic gain with limited resources. Implementing genomic selection (GS) could increase genetic gain, but optimization of GS is needed to account for these programs' unique challenges and advantages. Here, we used simulations to identify conditions under which genomic-assisted recurrent selection (GARS) would be more effective than phenotypic recurrent selection (PRS) in small new breeding programs. We compared genetic gain, cost per unit gain, genetic variance, and prediction accuracy of GARS (two or three cycles per year) vs. PRS (one cycle per year) assuming various breeding population sizes and trait genetic architectures. For oligogenic architecture, the maximum relative genetic gain advantage of GARS over PRS was 12-88%, which was observed only during the first few cycles. For the polygenic architecture, GARS provided maximum relative genetic gain advantage of 26-165%, and was always superior to PRS. Average prediction accuracy declines substantially after several cycles of selection, suggesting the prediction models should be updated regularly. Updating prediction models every year increased the genetic gain by up to 33-39% compared to no-update scenarios. For small populations and oligogenic traits, cost per unit gain was lower in PRS than GARS. However, with larger populations and polygenic traits cost per unit gain was up to 67% lower in GARS than PRS. Collectively, the simulations suggest that GARS could increase the genetic gain in small young breeding programs by accelerating the breeding cycles and enabling evaluation of larger populations.

Genomic Selection in Preliminary Yield Trials in a Winter Wheat Breeding Program

Genomic prediction (GP) is now routinely performed in crop plants to predict unobserved phenotypes. The use of predicted phenotypes to make selections is an active area of research. Here, we evaluate GP for predicting grain yield and compare genomic and phenotypic selection by tracking lines advanced. We examined four independent nurseries of F3:6 and F3:7 lines trialed at 6 to 10 locations each year. Yield was analyzed using mixed models that accounted for experimental design and spatial variations. Genotype-by-sequencing provided nearly 27,000 high-quality SNPs. Average genomic predictive ability, estimated for each year by randomly masking lines as missing in steps of 10% from 10 to 90%, and using the remaining lines from the same year as well as lines from other years in a training set, ranged from 0.23 to 0.55. The predictive ability estimated for a new year using the other years ranged from 0.17 to 0.28. Further, we tracked lines advanced based on phenotype from each of the four F3:6 nurseries. Lines with both above average genomic estimated breeding value (GEBV) and phenotypic value (BLUP) were retained for more years compared to lines with either above average GEBV or BLUP alone. The number of lines selected for advancement was substantially greater when predictions were made with 50% of the lines from the testing year added to the training set. Hence, evaluation of only 50% of the lines yearly seems possible. This study provides insights to assess and integrate genomic selection in breeding programs of autogamous crops.

Investigating Drought Tolerance in Chickpea Using Genome-Wide Association Mapping and Genomic Selection Based on Whole-Genome Resequencing Data

Drought tolerance is a complex trait that involves numerous genes. Identifying key causal genes or linked molecular markers can facilitate the fast development of drought tolerant varieties. Using a whole-genome resequencing approach, we sequenced 132 chickpea varieties and advanced breeding lines and found more than 144,000 single nucleotide polymorphisms (SNPs). We measured 13 yield and yield-related traits in three drought-prone environments of Western Australia. The genotypic effects were significant for all traits, and many traits showed highly significant correlations, ranging from 0.83 between grain yield and biomass to -0.67 between seed weight and seed emergence rate. To identify candidate genes, the SNP and trait data were incorporated into the SUPER genome-wide association study (GWAS) model, a modified version of the linear mixed model. We found that several SNPs from auxin-related genes, including auxin efflux carrier protein (PIN3), p-glycoprotein, and nodulin MtN21/EamA-like transporter, were significantly associated with yield and yield-related traits under drought-prone environments. We identified four genetic regions containing SNPs significantly associated with several different traits, which was an indication of pleiotropic effects. We also investigated the possibility of incorporating the GWAS results into a genomic selection (GS) model, which is another approach to deal with complex traits. Compared to using all SNPs, application of the GS model using subsets of SNPs significantly associated with the traits under investigation increased the prediction accuracies of three yield and yield-related traits by more than twofold. This has important implication for implementing GS in plant breeding programs.