Whole-genome Prediction Models Research Articles

Using genomic prediction with crop growth models enables the prediction of associated traits in wheat.

Crop growth models (CGM) can predict the performance of a cultivar in untested environments by sampling genotype-specific parameters. As they cannot predict the performance of new cultivars, it has been proposed to integrate CGMs with whole genome prediction (WGP) to combine the benefits of both models. Here, we used a CGM-WGP model to predict the performance of new wheat (Triticum aestivum) genotypes. The CGM was designed to predict phenology, nitrogen, and biomass traits. The CGM-WGP model simulated more heritable GSPs compared with the CGM and gave smaller errors for the observed phenotypes. The WGP model performed better when predicting yield, grain number, and grain protein content, but showed comparable performance to the CGM-WGP model for heading and physiological maturity dates. However, the CGM-WGP model was able to predict unobserved traits (for which there were no phenotypic records in the reference population). The CGM-WGP model also showed superior performance when predicting unrelated individuals that clustered separately from the reference population. Our results demonstrate new advantages for CGM-WGP modelling and suggest future efforts should focus on calibrating CGM-WGP models using high-throughput phenotypic measures that are cheaper and less laborious to collect.

High-resolution genome-wide association study pinpoints metal transporter and chelator genes involved in the genetic control of element levels in maize grain.

Despite its importance to plant function and human health, the genetics underpinning element levels in maize grain remain largely unknown. Through a genome-wide association study in the maize Ames panel of nearly 2,000 inbred lines that was imputed with ∼7.7 million SNP markers, we investigated the genetic basis of natural variation for the concentration of 11 elements in grain. Novel associations were detected for the metal transporter genes rte2 (rotten ear2) and irt1 (iron-regulated transporter1) with boron and nickel, respectively. We also further resolved loci that were previously found to be associated with one or more of five elements (copper, iron, manganese, molybdenum, and/or zinc), with two metal chelator and five metal transporter candidate causal genes identified. The nas5 (nicotianamine synthase5) gene involved in the synthesis of nicotianamine, a metal chelator, was found associated with both zinc and iron and suggests a common genetic basis controlling the accumulation of these two metals in the grain. Furthermore, moderate predictive abilities were obtained for the 11 elemental grain phenotypes with two whole-genome prediction models: Bayesian Ridge Regression (0.33–0.51) and BayesB (0.33–0.53). Of the two models, BayesB, with its greater emphasis on large-effect loci, showed ∼4–10% higher predictive abilities for nickel, molybdenum, and copper. Altogether, our findings contribute to an improved genotype-phenotype map for grain element accumulation in maize.

Nonlinear kernels, dominance, and envirotyping data increase the accuracy of genome-based prediction in multi-environment trials

Modern whole-genome prediction (WGP) frameworks that focus on multi-environment trials (MET) integrate large-scale genomics, phenomics, and envirotyping data. However, the more complex the statistical model, the longer the computational processing times, which do not always result in accuracy gains. We investigated the use of new kernel methods and modeling structures involving genomics and nongenomic sources of variation in two MET maize data sets. Five WGP models were considered, advancing in complexity from a main-effect additive model (A) to more complex structures, including dominance deviations (D), genotype × environment interaction (AE and DE), and the reaction-norm model using environmental covariables (W) and their interaction with A and D (AW + DW). A combination of those models built with three different kernel methods, Gaussian kernel (GK), Deep kernel (DK), and the benchmark genomic best linear-unbiased predictor (GBLUP/GB), was tested under three prediction scenarios: newly developed hybrids (CV1), sparse MET conditions (CV2), and new environments (CV0). GK and DK outperformed GB in prediction accuracy and reduction of computation time (~up to 20%) under all model–kernel scenarios. GK was more efficient in capturing the variation due to A + AE and D + DE effects and translated it into accuracy gains (~up to 85% compared with GB). DK provided more consistent predictions, even for more complex structures such as W + AW + DW. Our results suggest that DK and GK are more efficient in translating model complexity into accuracy, and more suitable for including dominance and reaction-norm effects in a biologically accurate and faster way.

Exploitation of Heterosis in Pearl Millet: A Review.

The phenomenon of heterosis has fascinated plant breeders ever since it was first described by Charles Darwin in 1876 in the vegetable kingdom and later elaborated by George H Shull and Edward M East in maize during 1908. Heterosis is the phenotypic and functional superiority manifested in the F1 crosses over the parents. Various classical complementation mechanisms gave way to the study of the underlying potential cellular and molecular mechanisms responsible for heterosis. In cereals, such as maize, heterosis has been exploited very well, with the development of many single-cross hybrids that revolutionized the yield and productivity enhancements. Pearl millet (Pennisetum glaucum (L.) R. Br.) is one of the important cereal crops with nutritious grains and lower water and energy footprints in addition to the capability of growing in some of the harshest and most marginal environments of the world. In this highly cross-pollinating crop, heterosis was exploited by the development of a commercially viable cytoplasmic male-sterility (CMS) system involving a three-lines breeding system (A-, B- and R-lines). The first set of male-sterile lines, i.e., Tift 23A and Tift18A, were developed in the early 1960s in Tifton, Georgia, USA. These provided a breakthrough in the development of hybrids worldwide, e.g., Tift 23A was extensively used by Punjab Agricultural University (PAU), Ludhiana, India, for the development of the first single-cross pearl millet hybrid, named Hybrid Bajra 1 (HB 1), in 1965. Over the past five decades, the pearl millet community has shown tremendous improvement in terms of cytoplasmic and nuclear diversification of the hybrid parental lines, which led to a progressive increase in the yield and adaptability of the hybrids that were developed, resulting in significant genetic gains. Lately, the whole genome sequencing of Tift 23D2B1 and re-sequencing of circa 1000 genomes by a consortium led by the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT) has been a significant milestone in the development of cutting-edge genetic and genomic resources in pearl millet. Recently, the application of genomics and molecular technologies has provided better insights into genetic architecture and patterns of heterotic gene pools. Development of whole-genome prediction models incorporating heterotic gene pool models, mapped traits and markers have the potential to take heterosis breeding to a new level in pearl millet. This review discusses advances and prospects in various fronts of heterosis for pearl millet.

Genome-Wide Association and Regional Heritability Mapping of Plant Architecture, Lodging and Productivity in Phaseolus vulgaris.

The availability of high-density molecular markers in common bean has allowed to explore the genetic basis of important complex agronomic traits with increased resolution. Genome-Wide Association Studies (GWAS) and Regional Heritability Mapping (RHM) are two analytical approaches for the detection of genetic variants. We carried out GWAS and RHM for plant architecture, lodging and productivity across two important growing environments in Brazil in a germplasm of 188 common bean varieties using DArTseq genotyping strategies. The coefficient of determination of G × E interaction (c2int) was equal to 17, 21 and 41%, respectively for the traits architecture, lodging, and productivity. Trait heritabilities were estimated at 0.81 (architecture), 0.79 (lodging) and 0.43 (productivity), and total genomic heritability accounted for large proportions (72% to ≈100%) of trait heritability. At the same probability threshold, three marker–trait associations were detected using GWAS, while RHM detected eight QTL encompassing 145 markers along five chromosomes. The proportion of genomic heritability explained by RHM was considerably higher (35.48 to 58.02) than that explained by GWAS (28.39 to 30.37). In general, RHM accounted for larger fractions of the additive genetic variance being captured by markers effects inside the defined regions. Nevertheless, a considerable proportion of the heritability is still missing (∼42% to ∼64%), probably due to LD between markers and genes and/or rare allele variants not sampled. RHM in autogamous species had the potential to identify larger-effect QTL combining allelic variants that could be effectively incorporated into whole-genome prediction models and tracked through breeding generations using marker-assisted selection.

Exploring new alleles for frost tolerance in winter rye.

Rye genetic resources provide a valuable source of new alleles for the improvement of frost tolerance in rye breeding programs. Frost tolerance is a must-have trait for winter cereal production in northern and continental cropping areas. Genetic resources should harbor promising alleles for the improvement of frost tolerance of winter rye elite lines. For frost tolerance breeding, the identification of quantitative trait loci (QTL) and the choice of optimum genome-based selection methods are essential. We identified genomic regions involved in frost tolerance of winter rye by QTL mapping in a biparental population derived from a highly frost tolerant selection from the Canadian cultivar Puma and the European elite line Lo157. Lines per se and their testcrosses were phenotyped in a controlled freeze test and in multi-location field trials in Russia and Canada. Three QTL on chromosomes 4R, 5R, and 7R were consistently detected across environments. The QTL on 5R is congruent with the genomic region harboring the Frost resistance locus 2 (Fr-2) in Triticeae. The Puma allele at the Fr-R2 locus was found to significantly increase frost tolerance. A comparison of predictive ability obtained from the QTL-based model with different whole-genome prediction models revealed that besides a few large, also small QTL effects contribute to the genomic variance of frost tolerance in rye. Genomic prediction models assigning a high weight to the Fr-R2 locus allow increasing the selection intensity for frost tolerance by genome-based pre-selection of promising candidates.

Limited influence of germline genetic variation on all-cause mortality in women with early onset breast cancer: evidence from gene-based tests, single-marker regression, and whole-genome prediction.

Women diagnosed with breast cancer have heterogeneous survival outcomes that cannot be fully explained by known prognostic factors, and germline variation is a plausible but unconfirmed risk factor. We used three approaches to test the hypothesis that germline variation drives some differences in survival: mortality loci identification, tumor aggressiveness loci identification, and whole-genome prediction. The 2954 study participants were women diagnosed with breast cancer before age 50, with a median follow-up of 15yearswho were genotyped on an exome array. We first searched for loci in gene regions that were associated with all-cause mortality. We next searched for loci in gene regions associated with five histopathological characteristics related to tumor aggressiveness. Last, we also predicted 10-year all-cause mortality on a subset of 1903 participants (3,245,343 variants after imputation) using whole-genome prediction methods. No risk loci for mortality or tumor aggressiveness were identified. This null result persisted when restricting to women with estrogen receptor-positive tumors, when examining suggestive loci in an independent study, and when restricting to previously published risk loci. Additionally, the whole-genome prediction model also found no evidence to support an association. Despite multiple complementary approaches, our study found no evidence that mortality in women with early onset breast cancer is influenced by germline variation.

Advances and Challenges in Genomic Selection for Disease Resistance

Breeding for disease resistance is a central focus of plant breeding programs, as any successful variety must have the complete package of high yield, disease resistance, agronomic performance, and end-use quality. With the need to accelerate the development of improved varieties, genomics-assisted breeding is becoming an important tool in breeding programs. With marker-assisted selection, there has been success in breeding for disease resistance; however, much of this work and research has focused on identifying, mapping, and selecting for major resistance genes that tend to be highly effective but vulnerable to breakdown with rapid changes in pathogen races. In contrast, breeding for minor-gene quantitative resistance tends to produce more durable varieties but is a more challenging breeding objective. As the genetic architecture of resistance shifts from single major R genes to a diffused architecture of many minor genes, the best approach for molecular breeding will shift from marker-assisted selection to genomic selection. Genomics-assisted breeding for quantitative resistance will therefore necessitate whole-genome prediction models and selection methodology as implemented for classical complex traits such as yield. Here, we examine multiple case studies testing whole-genome prediction models and genomic selection for disease resistance. In general, whole-genome models for disease resistance can produce prediction accuracy suitable for application in breeding. These models also largely outperform multiple linear regression as would be applied in marker-assisted selection. With the implementation of genomic selection for yield and other agronomic traits, whole-genome marker profiles will be available for the entire set of breeding lines, enabling genomic selection for disease at no additional direct cost. In this context, the scope of implementing genomics selection for disease resistance, and specifically for quantitative resistance and quarantined pathogens, becomes a tractable and powerful approach in breeding programs.

Genomic Selection for Processing and End-Use Quality Traits in the CIMMYT Spring Bread Wheat Breeding Program.

Wheat ( L.) cultivars must possess suitable end-use quality for release and consumer acceptability. However, breeding for quality traits is often considered a secondary target relative to yield largely because of amount of seed needed and expense. Without testing and selection, many undesirable materials are advanced, expending additional resources. Here, we develop and validate whole-genome prediction models for end-use quality phenotypes in the CIMMYT bread wheat breeding program. Model accuracy was tested using forward prediction on breeding lines ( = 5520) tested in unbalanced yield trials from 2009 to 2015 at Ciudad Obregon, Sonora, Mexico. Quality parameters included test weight, 1000-kernel weight, hardness, grain and flour protein, flour yield, sodium dodecyl sulfate sedimentation, Mixograph and Alveograph performance, and loaf volume. In general, prediction accuracy substantially increased over time as more data was available to train the model. Reflecting practical implementation of genomic selection (GS) in the breeding program, forward prediction accuracies () for quality parameters were assessed in 2015 and ranged from 0.32 (grain hardness) to 0.62 (mixing time). Increased selection intensity was possible with GS since more entries can be genotyped than phenotyped and expected genetic gain was 1.4 to 2.7 times higher across all traits than phenotypic selection. Given the limitations in measuring many lines for quality, we conclude that GS is a powerful tool to facilitate early generation selection for end-use quality in wheat, leaving larger populations for selection on yield during advanced testing and leading to better gain for both quality and yield in bread wheat breeding programs.

Incorporating parent-of-origin effects in whole-genome prediction of complex traits.

BackgroundParent-of-origin effects are due to differential contributions of paternal and maternal lineages to offspring phenotypes. Such effects include, for example, maternal effects in several species. However, epigenetically induced parent-of-origin effects have recently attracted attention due to their potential impact on variation of complex traits. Given that prediction of genetic merit or phenotypic performance is of interest in the study of complex traits, it is relevant to consider parent-of-origin effects in such predictions. We built a whole-genome prediction model that incorporates parent-of-origin effects by considering parental allele substitution effects of single nucleotide polymorphisms and gametic relationships derived from a pedigree (the POE model). We used this model to predict body mass index in a mouse population, a trait that is presumably affected by parent-of-origin effects, and also compared the prediction performance to that of a standard additive model that ignores parent-of-origin effects (the ADD model). We also used simulated data to assess the predictive performance of the POE model under various circumstances, in which parent-of-origin effects were generated by mimicking an imprinting mechanism.ResultsThe POE model did not predict better than the ADD model in the real data analysis, probably due to overfitting, since the POE model had far more parameters than the ADD model. However, when applied to simulated data, the POE model outperformed the ADD model when the contribution of parent-of-origin effects to phenotypic variation increased. The superiority of the POE model over the ADD model was up to 8 % on predictive correlation and 5 % on predictive mean squared error.ConclusionsThe simulation and the negative result obtained in the real data analysis indicated that, in order to gain benefit from the POE model in terms of prediction, a sizable contribution of parent-of-origin effects to variation is needed and such variation must be captured by the genetic markers fitted. Recent studies, however, suggest that most parent-of-origin effects stem from epigenetic regulation but not from a change in DNA sequence. Therefore, integrating epigenetic information with genetic markers may help to account for parent-of-origin effects in whole-genome prediction.

Statistical and Computational Challenges in Whole Genome Prediction and Genome-Wide Association Analyses for Plant and Animal Breeding

Whole genome prediction (WGP) modeling and genome-wide association (GWA) analyses are big data issues in agricultural quantitative genetics. Both areas require meaningful input from the statistical scholarly community in order to further improve the accuracy of prediction of genetic merit and inference on putative causal variants as well as improving the computational efficiency of existing methods and algorithms. These concerns have become increasingly critical as new sequencing technologies will only exacerbate current model dimensionality problems. We focus primarily on mixed model and hierarchical Bayesian analyses which have been most commonly pursued by animal and plant breeders for WGP thus far. We draw attention to our observation that many such previous analyses have not carefully inferred upon hyperparameters defined at the top levels of the Bayesian model hierarchy, but simply arbitrarily specify their values. We also reassess previous discussions on WGP model dimensionality, believing that useful data augmentation schemes utilized in various Markov Chain Monte Carlo (MCMC) schemes have led to a general misunderstanding that heavy-tailed or variable selection-based WGP models may be highly parameterized relative to more standard mixed model representations. Computational efficiency is addressed with respect to MCMC and competitive, albeit approximate, alternatives. Furthermore, GWA analyses are reassessed, encouraging a greater reliance on shrinkage-based inferences based on critically chosen priors, instead of potentially nonreproducible fixed effects P value-based inference.

An Integrated Approach to Empirical Bayesian Whole Genome Prediction Modeling

Computational efficiency is an increasing concern for whole genome prediction (WGP) based on denser genetic marker panels such that algorithms other than Markov Chain Monte Carlo (MCMC) warrant greater consideration, particularly for hierarchical models that flexibly confer either heavy-tailed (e.g., BayesA) or stochastic search and variable selection (SSVS) instead of Gaussian specifications on marker effect distributions. The expectation maximization (EM) algorithm is one attractive alternative; however, recently proposed hierarchical model implementations of EM have not addressed formal estimation of underlying hyperparameters even though their specifications are known to impact WGP accuracy. Furthermore, EM can be sensitive to starting values. We develop and explore the properties of an empirical Bayes strategy by conditioning EM implementations of BayesA or SSVS WGP models on marginal modal estimation of variance components and other key hyperparameters. These empirical Bayes implementations are compared against their MCMC counterparts for estimation of hyperparameters and WGP accuracy, both within the context of a simulation study and application to a loblolly pine dataset. In all cases, starting values were deemed to be important for EM-based estimates. Starting values based on MCMC posterior means were preferable, whereas those based on setting all marker effects equal to zero generally led to inferior performance. Nevertheless, a recently proposed regularization procedure was useful in alleviating the impact of starting values in the EM implementation of the SSVS model, as was modifying the expectation step in the BayesA model to be based on relative variances rather than on relative precisions.

Improving the computational efficiency of fully Bayes inference and assessing the effect of misspecification of hyperparameters in whole-genome prediction models.

BackgroundThe reliability of whole-genome prediction models (WGP) based on using high-density single nucleotide polymorphism (SNP) panels critically depends on proper specification of key hyperparameters. A currently popular WGP model labeled BayesB specifies a hyperparameter π, that is `loosely used to describe the proportion of SNPs that are in linkage disequilibrium (LD) with causal variants. The remaining markers are specified to be random draws from a Student t distribution with key hyperparameters being degrees of freedom v and scale s2.MethodsWe consider three alternative Markov chain Monte Carlo (MCMC) approaches based on the use of Metropolis-Hastings (MH) to estimate these key hyperparameters. The first approach, termed DFMH, is based on a previously published strategy for which s2 is drawn by a Gibbs step and v is drawn by a MH step. The second strategy, termed UNIMH, substitutes MH for Gibbs when drawing s2 and further collapses or marginalizes the full conditional density of v. The third strategy, termed BIVMH, is based on jointly drawing the two hyperparameters in a bivariate MH step. We also tested the effect of misspecification of s2 for its effect on accuracy of genomic estimated breeding values (GEBV), yet allowing for inference on the other hyperparameters.ResultsThe UNIMH and BIVMH strategies had significantly greater (P < 0.05) computational efficiencies for estimating v and s2 than DFMH in BayesA (π = 1) and BayesB implementations. We drew similar conclusions based on an analysis of the public domain heterogeneous stock mice data. We also determined significant drops (P < 0.01) in accuracies of GEBV under BayesA by overspecifying s2, whereas BayesB was more robust to such misspecifications. However, understating s2 was compensated by counterbalancing inferences on v in BayesA and BayesB, and on π in BayesB.ConclusionsSampling strategies based solely on MH updates of v and s2, and collapsed representations of full conditional densities can improve the computational efficiency of MCMC relative to the use of Gibbs updates. We believe that proper inferences on s2, v and π are vital to ensure that the accuracy of GEBV is maximized when using parametric WGP models.Electronic supplementary materialThe online version of this article (doi:10.1186/s12711-015-0092-x) contains supplementary material, which is available to authorized users.

Genomic selection: genome-wide prediction in plant improvement

Association analysis is used to measure relations between markers and quantitative trait loci (QTL). Their estimation ignores genes with small effects that trigger underpinning quantitative traits. By contrast, genome-wide selection estimates marker effects across the whole genome on the target population based on a prediction model developed in the training population (TP). Whole-genome prediction models estimate all marker effects in all loci and capture small QTL effects. Here, we review several genomic selection (GS) models with respect to both the prediction accuracy and genetic gain from selection. Phenotypic selection or marker-assisted breeding protocols can be replaced by selection, based on whole-genome predictions in which phenotyping updates the model to build up the prediction accuracy.

A comprehensive genetic approach for improving prediction of skin cancer risk in humans.

Prediction of genetic risk for disease is needed for preventive and personalized medicine. Genome-wide association studies have found unprecedented numbers of variants associated with complex human traits and diseases. However, these variants explain only a small proportion of genetic risk. Mounting evidence suggests that many traits, relevant to public health, are affected by large numbers of small-effect genes and that prediction of genetic risk to those traits and diseases could be improved by incorporating large numbers of markers into whole-genome prediction (WGP) models. We developed a WGP model incorporating thousands of markers for prediction of skin cancer risk in humans. We also considered other ways of incorporating genetic information into prediction models, such as family history or ancestry (using principal components, PCs, of informative markers). Prediction accuracy was evaluated using the area under the receiver operating characteristic curve (AUC) estimated in a cross-validation. Incorporation of genetic information (i.e., familial relationships, PCs, or WGP) yielded a significant increase in prediction accuracy: from an AUC of 0.53 for a baseline model that accounted for nongenetic covariates to AUCs of 0.58 (pedigree), 0.62 (PCs), and 0.64 (WGP). In summary, prediction of skin cancer risk could be improved by considering genetic information and using a large number of single-nucleotide polymorphisms (SNPs) in a WGP model, which allows for the detection of patterns of genetic risk that are above and beyond those that can be captured using family history. We discuss avenues for improving prediction accuracy and speculate on the possible use of WGP to prospectively identify individuals at high risk.

Comparison of whole-genome prediction models for traits with contrasting genetic architecture in a diversity panel of maize inbred lines

BackgroundThere is increasing empirical evidence that whole-genome prediction (WGP) is a powerful tool for predicting line and hybrid performance in maize. However, there is a lack of knowledge about the sensitivity of WGP models towards the genetic architecture of the trait. Whereas previous studies exclusively focused on highly polygenic traits, important agronomic traits such as disease resistances, nutrifunctional or climate adaptational traits have a genetic architecture which is either much less complex or unknown. For such cases, information about model robustness and guidelines for model selection are lacking. Here, we compared five WGP models with different assumptions about the distribution of the underlying genetic effects. As contrasting model traits, we chose three highly polygenic agronomic traits and three metabolites each with a major QTL explaining 22 to 30% of the genetic variance in a panel of 289 diverse maize inbred lines genotyped with 56,110 SNPs.ResultsWe found the five WGP models to be remarkable robust towards trait architecture with the largest differences in prediction accuracies ranging between 0.05 and 0.14 for the same trait, most likely as the result of the high level of linkage disequilibrium prevailing in elite maize germplasm. Whereas RR-BLUP performed best for the agronomic traits, it was inferior to LASSO or elastic net for the three metabolites. We found the approach of genome partitioning of genetic variance, first applied in human genetics, as useful in guiding the breeder which model to choose, if prior knowledge of the trait architecture is lacking.ConclusionsOur results suggest that in diverse germplasm of elite maize inbred lines with a high level of LD, WGP models differ only slightly in their accuracies, irrespective of the number and effects of QTL found in previous linkage or association mapping studies. However, small gains in prediction accuracies can be achieved if the WGP model is selected according to the genetic architecture of the trait. If the trait architecture is unknown e.g. for novel traits which only recently received attention in breeding, we suggest to inspect the distribution of the genetic variance explained by each chromosome for guiding model selection in WGP.