Genomic Prediction Models Research Articles

Genomic prediction models trained with historical records enable populating the German ex situ genebank bio-digital resource center of barley (Hordeum sp.) with information on resistances to soilborne barley mosaic viruses

Key messageGenomic prediction with special weight of major genes is a valuable tool to populate bio-digital resource centers.Phenotypic information of crop genetic resources is a prerequisite for an informed selection that aims to broaden the genetic base of the elite breeding pools. We investigated the potential of genomic prediction based on historical screening data of plant responses against the Barley yellow mosaic viruses for populating the bio-digital resource center of barley. Our study includes dense marker data for 3838 accessions of winter barley, and historical screening data of 1751 accessions for Barley yellow mosaic virus (BaYMV) and of 1771 accessions for Barley mild mosaic virus (BaMMV). Linear mixed models were fitted by considering combinations for the effects of genotypes, years, and locations. The best linear unbiased estimations displayed a broad spectrum of plant responses against BaYMV and BaMMV. Prediction abilities, computed as correlations between predictions and observed phenotypes of accessions, were low for the marker-assisted selection approach amounting to 0.42. In contrast, prediction abilities of genomic best linear unbiased predictions were high, with values of 0.62 for BaYMV and 0.64 for BaMMV. Prediction abilities of genomic prediction were improved by up to ~ 5% using W-BLUP, in which more weight is given to markers with significant major effects found by association mapping. Our results outline the utility of historical screening data and W-BLUP model to predict the performance of the non-phenotyped individuals in genebank collections. The presented strategy can be considered as part of the different approaches used in genebank genomics to valorize genetic resources for their usage in disease resistance breeding and research.

Comparison of Genomic Prediction Methods for Yellow, Stem, and Leaf Rust Resistance in Wheat Landraces from Afghanistan.

Wheat rust diseases, including yellow rust (Yr; also known as stripe rust) caused by Puccinia striiformis Westend. f. sp. tritici, leaf rust (Lr) caused by Puccinia triticina Eriks. and stem rust (Sr) caused by Puccinia graminis Pres f. sp. tritici are major threats to wheat production all around the globe. Durable resistance to wheat rust diseases can be achieved through genomic-assisted prediction of resistant accessions to increase genetic gain per unit time. Genomic prediction (GP) is a promising technology that uses genomic markers to estimate genomic-assisted breeding values (GBEVs) for selecting resistant plant genotypes and accumulating favorable alleles for adult plant resistance (APR) to wheat rust diseases. To evaluate GP we compared the predictive ability of nine different parametric, semi-parametric and Bayesian models including Genomic Unbiased Linear Prediction (GBLUP), Ridge Regression (RR), Least Absolute Shrinkage and Selection Operator (LASSO), Elastic Net (EN), Bayesian Ridge Regression (BRR), Bayesian A (BA), Bayesian B (BB), Bayesian C (BC) and Reproducing Kernel Hilbert Spacing model (RKHS) to estimate GEBV’s for APR to yellow, leaf and stem rust of wheat in a panel of 363 bread wheat landraces of Afghanistan origin. Based on five-fold cross validation the mean predictive abilities were 0.33, 0.30, 0.38, and 0.33 for Yr (2016), Yr (2017), Lr, and Sr, respectively. No single model outperformed the rest of the models for all traits. LASSO and EN showed the lowest predictive ability in four of the five traits. GBLUP and RR gave similar predictive abilities, whereas Bayesian models were not significantly different from each other as well. We also investigated the effect of the number of genotypes and the markers used in the analysis on the predictive ability of the GP model. The predictive ability was highest with 1000 markers and there was a linear trend in the predictive ability and the size of the training population. The results of the study are encouraging, confirming the feasibility of GP to be effectively applied in breeding programs for resistance to all three wheat rust diseases.

Utility of Climatic Information via Combining Ability Models to Improve Genomic Prediction for Yield Within the Genomes to Fields Maize Project.

Genomic prediction provides an efficient alternative to conventional phenotypic selection for developing improved cultivars with desirable characteristics. New and improved methods to genomic prediction are continually being developed that attempt to deal with the integration of data types beyond genomic information. Modern automated weather systems offer the opportunity to capture continuous data on a range of environmental parameters at specific field locations. In principle, this information could characterize training and target environments and enhance predictive ability by incorporating weather characteristics as part of the genotype-by-environment (G×E) interaction component in prediction models. We assessed the usefulness of including weather data variables in genomic prediction models using a naïve environmental kinship model across 30 environments comprising the Genomes to Fields (G2F) initiative in 2014 and 2015. Specifically four different prediction scenarios were evaluated (i) tested genotypes in observed environments; (ii) untested genotypes in observed environments; (iii) tested genotypes in unobserved environments; and (iv) untested genotypes in unobserved environments. A set of 1,481 unique hybrids were evaluated for grain yield. Evaluations were conducted using five different models including main effect of environments; general combining ability (GCA) effects of the maternal and paternal parents modeled using the genomic relationship matrix; specific combining ability (SCA) effects between maternal and paternal parents; interactions between genetic (GCA and SCA) effects and environmental effects; and finally interactions between the genetics effects and environmental covariates. Incorporation of the genotype-by-environment interaction term improved predictive ability across all scenarios. However, predictive ability was not improved through inclusion of naive environmental covariates in G×E models. More research should be conducted to link the observed weather conditions with important physiological aspects in plant development to improve predictive ability through the inclusion of weather data.

Genomic studies of milk-related traits in water buffalo (Bubalus bubalis) based on single-step genomic best linear unbiased prediction and random regression models

Genomic selection has been widely implemented in many livestock breeding programs, but it remains incipient in buffalo. Therefore, this study aimed to (1) estimate variance components incorporating genomic information in Murrah buffalo; (2) evaluate the performance of genomic prediction for milk-related traits using single- and multitrait random regression models (RRM) and the single-step genomic best linear unbiased prediction approach; and (3) estimate longitudinal SNP effects and candidate genes potentially associated with time-dependent variation in milk, fat, and protein yields, as well as somatic cell score (SCS) in multiple parities. The data used to estimate the genetic parameters consisted of a total of 323,140 test-day records. The average daily heritability estimates were moderate (0.35 ± 0.02 for milk yield, 0.22 ± 0.03 for fat yield, 0.42 ± 0.03 for protein yield, and 0.16 ± 0.03 for SCS). The highest heritability estimates, considering all traits studied, were observed between 20 and 280 d in milk (DIM). The genetic correlation estimates at different DIM among the evaluated traits ranged from -0.10 (156 to 185 DIM for SCS) to 0.61 (36 to 65 DIM for fat yield). In general, direct selection for any of the traits evaluated is expected to result in indirect genetic gains for milk yield, fat yield, and protein yield but also increase SCS at certain lactation stages, which is undesirable. The predicted RRM coefficients were used to derive the genomic estimated breeding values (GEBV) for each time point (from 5 to 305 DIM). In general, the tuning parameters evaluated when constructing the hybrid genomic relationship matrices had a small effect on the GEBV accuracy and a greater effect on the bias estimates. The SNP solutions were back-solved from the GEBV predicted from the Legendre random regression coefficients, which were then used to estimate the longitudinal SNP effects (from 5 to 305 DIM). The daily SNP effect for 3 different lactation stages were performed considering 3 different lactation stages for each trait and parity: from 5 to 70, from 71 to 150, and from 151 to 305 DIM. Important genomic regions related to the analyzed traits and parities that explain more than 0.50% of the total additive genetic variance were selected for further analyses of candidate genes. In general, similar potential candidate genes were found between traits, but our results suggest evidence of differential sets of candidate genes underlying the phenotypic expression of the traits across parities. These results contribute to a better understanding of the genetic architecture of milk production traits in dairy buffalo and reinforce the relevance of incorporating genomic information to genetically evaluate longitudinal traits in dairy buffalo. Furthermore, the candidate genes identified can be used as target genes in future functional genomics studies.

Feature Selection Stability and Accuracy of Prediction Models for Genomic Prediction of Residual Feed Intake in Pigs Using Machine Learning.

Feature selection (FS, i.e., selection of a subset of predictor variables) is essential in high-dimensional datasets to prevent overfitting of prediction/classification models and reduce computation time and resources. In genomics, FS allows identifying relevant markers and designing low-density SNP chips to evaluate selection candidates. In this research, several univariate and multivariate FS algorithms combined with various parametric and non-parametric learners were applied to the prediction of feed efficiency in growing pigs from high-dimensional genomic data. The objective was to find the best combination of feature selector, SNP subset size, and learner leading to accurate and stable (i.e., less sensitive to changes in the training data) prediction models. Genomic best linear unbiased prediction (GBLUP) without SNP pre-selection was the benchmark. Three types of FS methods were implemented: (i) filter methods: univariate (univ.dtree, spearcor) or multivariate (cforest, mrmr), with random selection as benchmark; (ii) embedded methods: elastic net and least absolute shrinkage and selection operator (LASSO) regression; (iii) combination of filter and embedded methods. Ridge regression, support vector machine (SVM), and gradient boosting (GB) were applied after pre-selection performed with the filter methods. Data represented 5,708 individual records of residual feed intake to be predicted from the animal’s own genotype. Accuracy (stability of results) was measured as the median (interquartile range) of the Spearman correlation between observed and predicted data in a 10-fold cross-validation. The best prediction in terms of accuracy and stability was obtained with SVM and GB using 500 or more SNPs [0.28 (0.02) and 0.27 (0.04) for SVM and GB with 1,000 SNPs, respectively]. With larger subset sizes (1,000–1,500 SNPs), the filter method had no influence on prediction quality, which was similar to that attained with a random selection. With 50–250 SNPs, the FS method had a huge impact on prediction quality: it was very poor for tree-based methods combined with any learner, but good and similar to what was obtained with larger SNP subsets when spearcor or mrmr were implemented with or without embedded methods. Those filters also led to very stable results, suggesting their potential use for designing low-density SNP chips for genome-based evaluation of feed efficiency.

Variance Component Decomposition for Growth Traits of the Bay Scallop (Argopecten irradians irradians)

The bay scallop (Argopecten irradians irradians) is one of the most important shellfish species in China. Since their introduction into China, only mass selection has been used in bay scallop breeding. With its gradual expansion and shortage of mate selection, population homozygosity increased, and fitness decreased. To investigate the effects of inbreeding and provide reference for improving breeding strategies and mating management, the variance components of the growth traits of the bay scallop were decomposed with genomic relationship matrices. The results indicated that the genetic variations in shell height and length were mainly accounted for by the additive effects. The genetic variation in shell width was mainly caused by dominance or dominance-by-dominance epistasis. The genetic variation in body weight was accounted for by dominance. No significant directional dominances were detected for all growth traits. Cross-validation for genomic prediction showed that including insignificant inbreeding in the genomic prediction model is not necessary, and we suggest that the genomic prediction model should be optimized with both likelihood ratio tests and cross-validation before utilization in practice.

Genetic Architecture and Genomic Prediction of Cooking Time in Common Bean (Phaseolus vulgaris L.).

Cooking time of the common bean is an important trait for consumer preference, with implications for nutrition, health, and environment. For efficient germplasm improvement, breeders need more information on the genetics to identify fast cooking sources with good agronomic properties and molecular breeding tools. In this study, we investigated a broad genetic variation among tropical germplasm from both Andean and Mesoamerican genepools. Four populations were evaluated for cooking time (CKT), water absorption capacity (WAC), and seed weight (SdW): a bi-parental RIL population (DxG), an eight-parental Mesoamerican MAGIC population, an Andean (VEF), and a Mesoamerican (MIP) breeding line panel. A total of 922 lines were evaluated in this study. Significant genetic variation was found in all populations with high heritabilities, ranging from 0.64 to 0.89 for CKT. CKT was related to the color of the seed coat, with the white colored seeds being the ones that cooked the fastest. Marker trait associations were investigated by QTL analysis and GWAS, resulting in the identification of 10 QTL. In populations with Andean germplasm, an inverse correlation of CKT and WAC, and also a QTL on Pv03 that inversely controls CKT and WAC (CKT3.2/WAC3.1) were observed. WAC7.1 was found in both Mesoamerican populations. QTL only explained a small part of the variance, and phenotypic distributions support a more quantitative mode of inheritance. For this reason, we evaluated how genomic prediction (GP) models can capture the genetic variation. GP accuracies for CKT varied, ranging from good results for the MAGIC population (0.55) to lower accuracies in the MIP panel (0.22). The phenotypic characterization of parental material will allow for the cooking time trait to be implemented in the active germplasm improvement programs. Molecular breeding tools can be developed to employ marker-assisted selection or genomic selection, which looks to be a promising tool in some populations to increase the efficiency of breeding activities.

Genomic and pedigree\u2010based predictive ability for quality traits in tea (Camellia sinensis (L.) O. Kuntze)

Genetic improvement of quality traits in tea (Camellia sinensis (L.) O. Kuntze) through conventional breeding methods has been limited, because tea quality is a difficult and expensive trait to measure. Genomic selection (GS) is suitable for predicting such complex traits, as it uses genome wide markers to estimate the genetic values of individuals. We compared the prediction accuracies of six genomic prediction models including Bayesian ridge regression (BRR), genomic best linear unbiased prediction (GBLUP), BayesA, BayesB, BayesC and reproducing kernel Hilbert spaces models incorporating the pedigree relationship namely; RKHS-pedigree, RKHS-markers and RKHS markers and pedigree (RKHS-MP) to determine the breeding values for 12 tea quality traits. One hundred and three tea genotypes were genotyped using genotyping-by-sequencing and phenotyped using nuclear magnetic resonance spectroscopy in replicated trials. We also compared the effect of trait heritability and training population size on prediction accuracies. The traits with the highest prediction accuracies were; theogallin (0.59), epicatechin gallate (ECG) (0.56) and theobromine (0.61), while the traits with the lowest prediction accuracies were theanine (0.32) and caffeine (0.39). The performance of all the GS models were almost the same, with BRR (0.53), BayesA (0.52), GBLUP (0.50) and RKHS-MP (0.50) performing slightly better than the others. Heritability estimates were moderate to high (0.35–0.92). Prediction accuracies increased with increasing training population size and trait heritability. We conclude that the moderate to high prediction accuracies observed suggests GS is a promising approach in tea improvement and could be implemented in breeding programmes.

Genome-wide association study of cyanogenic glycosides, proline, sugars, and pigments in Eucalyptus cladocalyx after 18 consecutive dry summers.

Natural variation of cyanogenic glycosides, soluble sugars, proline, and nondestructive optical sensing of pigments (chlorophyll, flavonols, and anthocyanins) was examined in ex situ natural populations of Eucalyptus cladocalyx F. Muell. grown under dry environmental conditions in the southern Atacama Desert, Chile. After 18 consecutive dry seasons, considerable plant-to-plant phenotypic variation for all the traits was observed in the field. For example, leaf hydrogen cyanide (HCN) concentrations varied from 0 (two acyanogenic individuals) to 1.54 mg cyanide g-1 DW. Subsequent genome-wide association study revealed associations with several genes with a known function in plants. HCN content was associated robustly with genes encoding Cytochrome P450 proteins, and with genes involved in the detoxification mechanism of HCN in cells (β-cyanoalanine synthase and cyanoalanine nitrilase). Another important finding was that sugars, proline, and pigment content were linked to genes involved in transport, biosynthesis, and/or catabolism. Estimates of genomic heritability (based on haplotypes) ranged between 0.46 and 0.84 (HCN and proline content, respectively). Proline and soluble sugars had the highest predictive ability of genomic prediction models (PA=0.65 and PA=0.71, respectively). PA values for HCN content and flavonols were relatively moderate, with estimates ranging from 0.44 to 0.50. These findings provide new understanding on the genetic architecture of cyanogenic capacity, and other key complex traits in cyanogenic E. cladocalyx.

Heuristic hyperparameter optimization of deep learning models for genomic prediction.

There is a growing interest among quantitative geneticists and animal breeders in the use of deep learning (DL) for genomic prediction. However, the performance of DL is affected by hyperparameters that are typically manually set by users. These hyperparameters do not simply specify the architecture of the model; they are also critical for the efficacy of the optimization and model-fitting process. To date, most DL approaches used for genomic prediction have concentrated on identifying suitable hyperparameters by exploring discrete options from a subset of the hyperparameter space. Enlarging the hyperparameter optimization search space with continuous hyperparameters is a daunting combinatorial problem. To deal with this problem, we propose using differential evolution (DE) to perform an efficient search of arbitrarily complex hyperparameter spaces in DL models, and we apply this to the specific case of genomic prediction of livestock phenotypes. This approach was evaluated on two pig and cattle datasets with real genotypes and simulated phenotypes (N = 7,539 animals and M = 48,541 markers) and one real dataset (N = 910 individuals and M = 28,916 markers). Hyperparameters were evaluated using cross-validation. We compared the predictive performance of DL models using hyperparameters optimized by DE against DL models with “best practice” hyperparameters selected from published studies and baseline DL models with randomly specified hyperparameters. Optimized models using DE showed a clear improvement in predictive performance across all three datasets. DE optimized hyperparameters also resulted in DL models with less overfitting and less variation in predictive performance over repeated retraining compared to non-optimized DL models.

Translating insights from the seed metabolome into improved prediction for lipid-composition traits in oat (Avena sativa L.).

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.

Evaluating the accuracy of genomic prediction for the management and conservation of relictual natural tree populations

Studying and understanding the evolution of relictual natural populations is critical for developing conservation initiatives of endangered species, such as management in situ and assisted migration. Recently, genomic and bioinformatics tools have promised a wide avenue for developing more efficient programs. Genomic prediction (GP) models are one of such tools; although, in trees, only some successful examples exit. They have mostly been used to increase predictive ability in commercial traits and reduce breeding cycle length. Thus, it remains to be tested whether GP can be extended for the management and conservation of natural small and secluded populations. Here, we explored such a possibility in a pilot study to predict the performance of introduced saplings in a managed population of sacred fir (Abies religiosa; Pinaceae) in central Mexico. We genotyped over 200 naturally re-generated and introduced individuals with 2286 single nucleotide polymorphisms (SNP), derived from genotyping by sequencing, and used them to develop GP models for growth and physiological traits. After testing different training and validation datasets, and determining predictive ability of “across-groups” models with cross-validation techniques, acceptable predictive abilities (ry) were obtained for growth during the previous growing season, water potential, stem diameter, and aboveground biomass (0.36, 0.27, 0.26, and 0.24, respectively). The best models were always those built with natural saplings and used to predict the early performance of introduced individuals in the same environment, although fair predictabilities were also obtained when predicting performance between natural populations. Model fine-tuning resulted in reduced datasets of approximately 700 SNPs that helped optimizing phenotype predictability, particularly for water potential, for which ry was up to 0.28. These pilot-scale results are preliminary but encouraging and justify additional research efforts for implementing GP in small and secluded natural populations, particularly for endangered non-model species.

Genomic prediction using training population design in interspecific soybean populations.

The online version contains supplementary material available at 10.1007/s11032-021-01203-6.

Independent Validation of Genomic Prediction in Strawberry Over Multiple Cycles.

The University of Florida strawberry (Fragaria × ananassa) breeding program has implemented genomic prediction (GP) as a tool for choosing outstanding parents for crosses over the last five seasons. This has allowed the use of some parents 1 year earlier than with traditional methods, thus reducing the duration of the breeding cycle. However, as the number of breeding cycles increases over time, greater knowledge is needed on how multiple cycles can be used in the practical implementation of GP in strawberry breeding. Advanced selections and cultivars totaling 1,558 unique individuals were tested in field trials for yield and fruit quality traits over five consecutive years and genotyped for 9,908 SNP markers. Prediction of breeding values was carried out using Bayes B models. Independent validation was carried out using separate trials/years as training (TRN) and testing (TST) populations. Single-trial predictive abilities for five polygenic traits averaged 0.35, which was reduced to 0.24 when individuals common across trials were excluded, emphasizing the importance of relatedness among training and testing populations. Training populations including up to four previous breeding cycles increased predictive abilities, likely due to increases in both training population size and relatedness. Predictive ability was also strongly influenced by heritability, but less so by changes in linkage disequilibrium and effective population size. Genotype by year interactions were minimal. A strategy for practical implementation of GP in strawberry breeding is outlined that uses multiple cycles to predict parental performance and accounts for traits not included in GP models when constructing crosses. Given the importance of relatedness to the success of GP in strawberry, future work could focus on the optimization of relatedness in the design of TRN and TST populations to increase predictive ability in the short-term without compromising long-term genetic gains.

Optimising accuracy of performance predictions using available morphophysiological information in wheat breeding germplasm

AbstractWheat breeding programmes often harbour large number of developed progenies. Testing of all progenies in many target environments is not considered cost‐effective, so the genomic predictions are employed. Genomic predictions might be enhanced with adaptational traits given the cost‐effectiveness of their phenotyping. The aim of this study was to assess effectiveness of adding a priori available phenotyping data of adaptational morpho‐physiological traits to genomic predictions in Bayesian framework. The panel of 120 winter wheat genotypes was sown over four growing seasons (2013/2014–2016/2017) in field conditions in two replicates and phenotyped for target traits grain yield and plant height along with nine distinctness, uniformity and stability (DUS) traits. Five genotype groups were defined by the K‐means clustering method based on nine DUS traits. DUS traits were used as fixed covariates in five different genomic prediction models with different assumptions about the distribution densities of marker effects: Bayes A, Bayes B, Bayes C, Bayesian ridge regression and Bayesian lasso. Adding DUS traits as covariates significantly improved accuracies and reduced the root mean square errors (RMSEP) of genomic predictions. Marginal differences between different models were observed. Adding covariates to genomic prediction models might be good strategy to improve accuracies of the predictions by accounting for environmental adaptations, provided their a priori availability or low costs of additional phenotyping.

The importance of dominance and genotype-by-environment interactions on grain yield variation in a large-scale public cooperative maize experiment.

High-dimensional and high-throughput genomic, field performance, and environmental data are becoming increasingly available to crop breeding programs, and their integration can facilitate genomic prediction within and across environments and provide insights into the genetic architecture of complex traits and the nature of genotype-by-environment interactions. To partition trait variation into additive and dominance (main effect) genetic and corresponding genetic-by-environment variances, and to identify specific environmental factors that influence genotype-by-environment interactions, we curated and analyzed genotypic and phenotypic data on 1918 maize (Zea mays L.) hybrids and environmental data from 65 testing environments. For grain yield, dominance variance was similar in magnitude to additive variance, and genetic-by-environment variances were more important than genetic main effect variances. Models involving both additive and dominance relationships best fit the data and modeling unique genetic covariances among all environments provided the best characterization of the genotype-by-environment interaction patterns. Similarity of relative hybrid performance among environments was modeled as a function of underlying weather variables, permitting identification of weather covariates driving correlations of genetic effects across environments. The resulting models can be used for genomic prediction of mean hybrid performance across populations of environments tested or for environment-specific predictions. These results can also guide efforts to incorporate high-throughput environmental data into genomic prediction models and predict values in new environments characterized with the same environmental characteristics.

Optimization of training sets for genomic prediction of early-stage single crosses in maize.

Training population optimization algorithms are useful for efficiently training genomic prediction models for single-cross performance, especially if the population is extended beyond only realized crosses to all possible single crosses. Genomic prediction of single-cross performance could allow effective evaluation of all possible single crosses between all inbreds developed in a hybrid breeding program. The objectives of the present study were to investigate the effect of different levels of relatedness on genomic predictive ability of single crosses, evaluate the usefulness of deterministic formula to forecast prediction accuracy in advance, and determine the potential for TRS optimization based on prediction error variance (PEVmean) and coefficient of determination (CDmean) criteria. We used 481 single crosses made by crossing 89 random recombinant inbred lines (RILs) belonging to the Iowa stiff stalk synthetic group with 103 random RILs belonging to the non-stiff stalk synthetic heterotic group. As expected, predictive ability was enhanced by ensuring close relationships between TRSs and target sets, even when TRS sizes were smaller. We found that designing a TRS based on PEVmean or CDmean criteria is useful for increasing the efficiency of genomic prediction of maize single crosses. We went further and extended the sampling space from that of all observed single crosses to all possible single crosses, providing a much larger genetic space within which to design a training population. Using all possible single crosses increased the advantage of the PEVmean and CDmean methods based on expected prediction accuracy. This finding suggests that it may be worthwhile using an optimization algorithm to select a training population from all possible single crosses to maximize efficiency in training accurate models for hybrid genomic prediction.

A combined BSA-Seq and linkage mapping approach identifies genomic regions associated with Phytophthora root and crown rot resistance in squash

Two QTL mapping approaches were used to identify a total of six QTL associated with Phytophthora root and crown rot resistance in a biparental squash population. Phytophthora root and crown rot, caused by the soilborne oomycete pathogen Phytophthora capsici, leads to severe yield losses in squash (Cucurbita pepo). To identify quantitative trait loci (QTL) involved in resistance to this disease, we crossed a partially resistant squash breeding line with a susceptible zucchini cultivar and evaluated over 13,000 F2 seedlings in a greenhouse screen. Bulked segregant analysis with whole genome resequencing (BSA-Seq) resulted in the identification of five genomic regions-on chromosomes 4, 5, 8, 12, and 16-featuring significant allele frequency differentiation between susceptible and resistant bulks in each of two independent replicates. In addition, we conducted linkage mapping using a population of 176 F3 families derived from individually genotyped F2 individuals. Variation in disease severity among these families was best explained by a four-QTL model, comprising the same loci identified via BSA-Seq on chromosomes 4, 5, and 8 as well as an additional locus on chromosome 19, for a combined total of six QTL identified between both methods. Loci, whether those identified by BSA-Seq or linkage mapping, were of small-to-moderate effect, collectively accounting for 28-35% and individually for 2-10% of the phenotypic variance explained. However, a multiple linear regression model using one marker in each BSA-Seq QTL could predict F2:3 disease severity with only a slight drop in cross-validation accuracy compared to genomic prediction models using genome-wide markers. These results suggest that marker-assisted selection could be a suitable approach for improving Phytophthora crown and root rot resistance in squash.

Genomic selection in American mink (Neovison vison) using a single-step genomic best linear unbiased prediction model for size and quality traits graded on live mink.

Genomic selection relies on single-nucleotide polymorphisms (SNPs), which are often collected using medium-density SNP arrays. In mink, no such array is available; instead, genotyping by sequencing (GBS) can be used to generate marker information. Here, we evaluated the effect of genomic selection for mink using GBS. We compared the estimated breeding values (EBVs) from single-step genomic best linear unbiased prediction (SSGBLUP) models to the EBV from ordinary pedigree-based BLUP models. We analyzed seven size and quality traits from the live grading of brown mink. The phenotype data consisted of ~20,600 records for the seven traits from the mink born between 2013 and 2016. Genotype data included 2,103 mink born between 2010 and 2014, mostly breeding animals. In total, 28,336 SNP markers from 391 scaffolds were available for genomic prediction. The pedigree file included 29,212 mink. The predictive ability was assessed by the correlation (r) between progeny trait deviation (PTD) and EBV, and the regression of PTD on EBV, using 5-fold cross-validation. For each fold, one-fifth of animals born in 2014 formed the validation set. For all traits, the SSGBLUP model resulted in higher accuracies than the BLUP model. The average increase in accuracy was 15% (between 3% for fur clarity and 28% for body weight). For three traits (body weight, silky appearance of the under wool, and guard hair thickness), the difference in r between the two models was significant (P < 0.05). For all traits, the regression slopes of PTD on EBV from SSGBLUP models were closer to 1 than regression slopes from BLUP models, indicating SSGBLUP models resulted in less bias of EBV for selection candidates than the BLUP models. However, the regression coefficients did not differ significantly. In conclusion, the SSGBLUP model is superior to conventional BLUP model in the accurate selection of superior animals, and, thus, it would increase genetic gain in a selective breeding program. In addition, this study shows that GBS data work well in genomic prediction in mink, demonstrating the potential of GBS for genomic selection in livestock species.

The 4.(n) wheat breeding era: genomic-based predictions becoming part of the daily routine and beyond

Wheat (Triticum aestivum L.) is one of the most important crops worldwide, also as a role model in plant breeding and genomics. Along with gene edition, genomic selection (GS) is the most advancing area in plant breeding, and several developments are in curse in this field, which soon will result in the recently termed “breeding 4.0” stage. The aim of this review is to present and discuss the most recent and impacting advances in GS applied to wheat breeding. Wheat shows particular features, e.g. a crop with narrow genetic variability and a large and complex genome, which motivates especial discussions on the present theme. The advances in enviromics and phenomics are presented and also the way they literally enter the genomic prediction models. The most breeder-advantageous wheat genotyping platforms currently available are also presented. Regarding data analysis within the genomic selection scheme, machine learning and deep learning methods are the most advancing approaches for predicting phenotypes, and improvements on these algorithms are nowadays in the center of the debate. Several advances in course will move plant breeding from the current 3.0 to the 4.0 stage, so achieving the status of “rocket science” and requiring highly skilled breeders and multidisciplinary teams.