Multi-trait Genomic Prediction Research Articles

Multivariate genomic prediction for commercial traits of economic importance in Banana shrimp Fenneropenaeus merguiensis

Advantages of multi-trait machine and deep learning genomic prediction models for quantitative complex traits have not been documented or very limited in aquaculture species. Thus, the present study sought to understand effects of the multi-trait single-step genomic best linear unbiased prediction (ssGBLUP), Bayesian (BayesCpi), random forest (RF) and multilayer perceptron (MLP) models on genomic prediction accuracies for traits of commercial importance in banana white shrimp (Fenneropenaeus merguiensis). Our analyses were conducted in a breeding shrimp population comprising 562 individuals (offspring of 48 parental pairs) genotyped for 9472 single nucleotide polymorphisms (SNPs) and the animals had full phenotype records for five important traits (i.e., body weight, abdominal width, tail weight, raw colour of live shrimp and resistance to hepatopancreatic parvovirus). In both univariate and multi-trait analyses, machine (RF) and deep learning (MLP) models outperformed ssGBLUP for all traits studied. However, they had similar predictive performance to BayesCpi. The benefits of the multivariate relative to univariate models were trait- and method-specific. Multi-trait BayesCpi increased the prediction accuracies for growth (weight and width), carcass (tail weight) and HPV resistance by 9.3 to 17.8%. However, the multi-trait random forest models improved the predictive power for only abdominal width (14.3%) and disease resistance to hepatopancreatic parvovirus (10.0%). When the multi-trait MLP was used, the improvements in the prediction accuracies were observed for abdominal width and raw colour (4.9 and 6.0%, respectively). There were almost no differences in the predictive power between univariate and multi-trait ssGBLUP. Among the multi-trait models used, BayesCpi outperformed other methods (ssGBLUP, RF and MLP). It is concluded that either BayesCpi or machine and deep learning-based multi-trait genomic prediction models should be employed in large-scale genetic enhancement programs for banana shrimp. These approaches show enormous potential to enhance genetic progress made in this population of banana shrimp and potentially for other aquaculture species.

Multi-trait genomic selection can increase selection accuracy for deoxynivalenol accumulation resulting from fusarium head blight in wheat.

Multi-trait genomic prediction (MTGP) can improve selection accuracy for economically valuable 'primary' traits by incorporating data on correlated secondary traits. Resistance to Fusarium head blight (FHB), a fungal disease of wheat (Triticum aestivum L.) and barley (Hordeum vulgare L.), is evaluated using four genetically correlated traits: incidence (INC), severity (SEV), Fusarium damaged kernels (FDK), and deoxynivalenol content (DON). Both FDK and DON are primary traits; DON evaluation is expensive and usually requires several months for wheat breeders to get results from service laboratories performing the evaluations. We evaluated MTGP for DON using three soft red winter wheat breeding datasets: two diversity panels from the University of Illinois (IL) and Purdue University (PU) and a dataset consisting of 2019-2020 University of Illinois breeding cohorts. For DON, relative to single-trait (ST) genomic prediction, MTGP including phenotypic data for secondary traits on both validation and training sets, resulted in 23.4 and 10.6% higher predictive abilities in IL and PU panels, respectively. The MTGP models were advantageous only when secondary traits were included in both training and validation sets. In addition, MTGP models were more accurate than ST models only when FDK was included, and once FDK was included in the model, adding additional traits hardly improved accuracy. Evaluation of MTGP models across testing cohorts indicated that MTGP could increase accuracy by more than twofold in the early stages. Overall, we show that MTGP can increase selection accuracy for resistance to DON accumulation in wheat provided FDK is evaluated on the selection candidates.

MegaLMM: Mega-scale linear mixed models for genomic predictions with thousands of traits

Large-scale phenotype data can enhance the power of genomic prediction in plant and animal breeding, as well as human genetics. However, the statistical foundation of multi-trait genomic prediction is based on the multivariate linear mixed effect model, a tool notorious for its fragility when applied to more than a handful of traits. We present MegaLMM, a statistical framework and associated software package for mixed model analyses of a virtually unlimited number of traits. Using three examples with real plant data, we show that MegaLMM can leverage thousands of traits at once to significantly improve genetic value prediction accuracy.

A Multiple-Trait Bayesian Variable Selection Regression Method for Integrating Phenotypic Causal Networks in Genome-Wide Association Studies

Bayesian regression methods that incorporate different mixture priors for marker effects are used in multi-trait genomic prediction. These methods can also be extended to genome-wide association studies (GWAS). In multiple-trait GWAS, incorporating the underlying causal structures among traits is essential for comprehensively understanding the relationship between genotypes and traits of interest. Therefore, we develop a GWAS methodology, SEM-Bayesian alphabet, which, by applying the structural equation model (SEM), can be used to incorporate causal structures into multi-trait Bayesian regression methods. SEM-Bayesian alphabet provides a more comprehensive understanding of the genotype-phenotype mapping than multi-trait GWAS by performing GWAS based on indirect, direct and overall marker effects. The superior performance of SEM-Bayesian alphabet was demonstrated by comparing its GWAS results with other similar multi-trait GWAS methods on real and simulated data. The software tool JWAS offers open-source routines to perform these analyses.

Multi-Trait Genomic Prediction of Yield-Related Traits in US Soft Wheat under Variable Water Regimes.

The performance of genomic prediction (GP) on genetically correlated traits can be improved through an interdependence multi-trait model under a multi-environment context. In this study, a panel of 237 soft facultative wheat (Triticum aestivum L.) lines was evaluated to compare single- and multi-trait models for predicting grain yield (GY), harvest index (HI), spike fertility (SF), and thousand grain weight (TGW). The panel was phenotyped in two locations and two years in Florida under drought and moderately drought stress conditions, while the genotyping was performed using 27,957 genotyping-by-sequencing (GBS) single nucleotide polymorphism (SNP) makers. Five predictive models including Multi-environment Genomic Best Linear Unbiased Predictor (MGBLUP), Bayesian Multi-trait Multi-environment (BMTME), Bayesian Multi-output Regressor Stacking (BMORS), Single-trait Multi-environment Deep Learning (SMDL), and Multi-trait Multi-environment Deep Learning (MMDL) were compared. Across environments, the multi-trait statistical model (BMTME) was superior to the multi-trait DL model for prediction accuracy in most scenarios, but the DL models were comparable to the statistical models for response to selection. The multi-trait model also showed 5 to 22% more genetic gain compared to the single-trait model across environment reflected by the response to selection. Overall, these results suggest that multi-trait genomic prediction can be an efficient strategy for economically important yield component related traits in soft wheat.

Multitrait genomic prediction of methane emissions in Danish Holstein cattle

In dairy cattle, selecting for lower methane-emitting animals is one of the new challenges of this decade. However, genetic selection requires a large number of animals with records to get accurate estimated breeding values (EBV). Given that CH4 records are scarce, the use of information on routinely recorded and highly correlated traits with CH4 has been suggested to increase the accuracy of genomic EBV (GEBV) through multitrait (genomic) prediction. Therefore, the objective of this study was to evaluate accuracies of prediction of GEBV for CH4 by including or omitting CH4, energy-corrected milk (ECM), and body weight (BW) as well as genotypic information in multitrait analyses across 2 methods: BLUP and single-step genomic BLUP (SSGBLUP). A total of 2,725 cows with CH4 concentration in breath (14,125 records), BW (61,667 records), and ECM (61,610 records) were included in the analyses. Approximately 2,000 of these cows were genotyped or imputed to 50K. Ten cross-validation groups were formed by randomly grouping paternal half-sibs. Five scenarios were performed: (1) base scenario with only CH4 information; (2) without CH4, but with information from BW, ECM, or BW+ECM only in reference population; (3) without CH4, but with information from BW, ECM, or BW+ECM in both validation and reference population; (4) with CH4 information and BW, ECM, or BW+ECM information only in the reference population; and (5) with CH4 information and BW, ECM, or BW+ECM information in both validation and reference population. As a result, for each method (BLUP, SSGBLUP), 13 sub-scenarios were performed, 1 from scenario 1, and 3 for each of the subsequent 4 scenarios. The average accuracy of GEBV for CH4 in the base scenario was 0.32 for BLUP and 0.42 for SSGBLUP, and it ranged from 0.10 in scenario 2 to 0.78 in scenario 5 across methods. In terms of bias, the base scenario 1 was unbiased for SSGBLUP; similar results were achieved with scenario 5. Including information on ECM increased the accuracy of GEBV for CH4 by up to 61%, whereas adding information on both traits (BW and ECM) increased the accuracy by up to 90%. Scenarios that did not include CH4 in the reference population had the lowest correlations (0.17-0.33) with single-trait CH4 GEBV, and scenarios with CH4 in the reference population had the highest correlations (0.41-0.81). Thus, failure to include CH4 in future reference populations results in predicted CH4 GEBV, which cannot be used in practical selection. Therefore, recording CH4 in more animals remains a priority. Finally, multiple-trait genomic prediction using routinely recorded BW and ECM leads to higher prediction accuracies than traditional single-trait genomic prediction for CH4 and is a viable solution for increasing the accuracies of GEBV for scarcely recorded CH4 in practice.

Prediction Strategies for Leveraging Information of Associated Traits under Single- and Multi-Trait Approaches in Soybeans

The availability of molecular markers has revolutionized conventional ways to improve genotypes in plant and animal breeding through genome-based predictions. Several models and methods have been developed to leverage the genomic information in the prediction context to allow more efficient ways to screen and select superior genotypes. In plant breeding, usually, grain yield (yield) is the main trait to drive the selection of superior genotypes; however, in many cases, the information of associated traits is also routinely collected and it can potentially be used to enhance the selection. In this research, we considered different prediction strategies to leverage the information of the associated traits ([AT]; full: all traits observed for the same genotype; and partial: some traits observed for the same genotype) under an alternative single-trait model and the multi-trait approach. The alternative single-trait model included the information of the AT for yield prediction via the phenotypic covariances while the multi-trait model jointly analyzed all the traits. The performance of these strategies was assessed using the marker and phenotypic information from the Soybean Nested Association Mapping (SoyNAM) project observed in Nebraska in 2012. The results showed that the alternative single-trait strategy, which combines the marker and the information of the AT, outperforms the multi-trait model by around 12% and the conventional single-trait strategy (baseline) by 25%. When no information on the AT was available for those genotypes in the testing sets, the multi-trait model reduced the baseline results by around 6%. For the cases where genotypes were partially observed (i.e., some traits observed but not others for the same genotype), the multi-trait strategy showed improvements of around 6% for yield and between 2% to 9% for the other traits. Hence, when yield drives the selection of superior genotypes, the single-trait and multi-trait genomic prediction will achieve significant improvements when some genotypes have been fully or partially tested, with the alternative single-trait model delivering the best results. These results provide empirical evidence of the usefulness of the AT for improving the predictive ability of prediction models for breeding applications.

Multi-trait Genomic Prediction Model Increased the Predictive Ability for Agronomic and Malting Quality Traits in Barley (Hordeum vulgare L.).

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

Combining Crop Growth Modeling and Statistical Genetic Modeling to Evaluate Phenotyping Strategies.

Genomic prediction of complex traits, say yield, benefits from including information on correlated component traits. Statistical criteria to decide which yield components to consider in the prediction model include the heritability of the component traits and their genetic correlation with yield. Not all component traits are easy to measure. Therefore, it may be attractive to include proxies to yield components, where these proxies are measured in (high-throughput) phenotyping platforms during the growing season. Using the Agricultural Production Systems Simulator (APSIM)-wheat cropping systems model, we simulated phenotypes for a wheat diversity panel segregating for a set of physiological parameters regulating phenology, biomass partitioning, and the ability to capture environmental resources. The distribution of the additive quantitative trait locus effects regulating the APSIM physiological parameters approximated the same distribution of quantitative trait locus effects on real phenotypic data for yield and heading date. We use the crop growth model APSIM-wheat to simulate phenotypes in three Australian environments with contrasting water deficit patterns. The APSIM output contained the dynamics of biomass and canopy cover, plus yield at the end of the growing season. Each water deficit pattern triggered different adaptive mechanisms and the impact of component traits differed between drought scenarios. We evaluated multiple phenotyping schedules by adding plot and measurement error to the dynamics of biomass and canopy cover. We used these trait dynamics to fit parametric models and P-splines to extract parameters with a larger heritability than the phenotypes at individual time points. We used those parameters in multi-trait prediction models for final yield. The combined use of crop growth models and multi-trait genomic prediction models provides a procedure to assess the efficiency of phenotyping strategies and compare methods to model trait dynamics. It also allows us to quantify the impact of yield components on yield prediction accuracy even in different environment types. In scenarios with mild or no water stress, yield prediction accuracy benefitted from including biomass and green canopy cover parameters. The advantage of the multi-trait model was smaller for the early-drought scenario, due to the reduced correlation between the secondary and the target trait. Therefore, multi-trait genomic prediction models for yield require scenario-specific correlated traits.

Pitfalls and Remedies for Cross Validation with Multi-trait Genomic Prediction Methods.

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

Multi-trait single-step genomic prediction accounting for heterogeneous (co)variances over the genome

Widely used genomic prediction models may not properly account for heterogeneous (co)variance structure across the genome. Models such as BayesA and BayesB assume locus-specific variance, which are highly influenced by the prior for (co)variance of single nucleotide polymorphism (SNP) effect, regardless of the size of data. Models such as BayesC or GBLUP assume a common (co)variance for a proportion (BayesC) or all (GBLUP) of the SNP effects. In this study, we propose a multi-trait Bayesian whole genome regression method (BayesN0), which is based on grouping a number of predefined SNPs to account for heterogeneous (co)variance structure across the genome. This model was also implemented in single-step Bayesian regression (ssBayesN0). For practical implementation, we considered multi-trait single-step SNPBLUP models, using (co)variance estimates from BayesN0 or ssBayesN0. Genotype data were simulated using haplotypes on first five chromosomes of 2200 Danish Holstein cattle, and phenotypes were simulated for two traits with heritabilities 0.1 or 0.4, assuming 200 quantitative trait loci (QTL). We compared prediction accuracy from different prediction models and different region sizes (one SNP, 100 SNPs, one chromosome or whole genome). In general, highest accuracies were obtained when 100 adjacent SNPs were grouped together. The ssBayesN0 improved accuracies over BayesN0, and using (co)variance estimates from ssBayesN0 generally yielded higher accuracies than using (co)variance estimates from BayesN0, for the 100 SNPs region size. Our results suggest that it could be a good strategy to estimate (co)variance components from ssBayesN0, and then to use those estimates in genomic prediction using multi-trait single-step SNPBLUP, in routine genomic evaluations.

Multi-Trait and Trait-Assisted Genomic Prediction of Winter Wheat Quality Traits Using Advanced Lines from Four Breeding Cycles

Genetically correlated traits can be used for improving predictive abilities of genomic predictions including several traits in multi-trait models. Here, the wheat quality traits thousand-kernel weight, grain protein content, Zeleny sedimentation, and falling number were phenotyped in 1152 advanced winter wheat lines from four cycles of a commercial breeding program. Multi-trait and trait-assisted genomic prediction models including two or four traits were studied and compared with single-trait models. In the trait-assisted genomic predictions, breeding values of the trait of interest were predicted for lines that had been phenotyped for additional traits. Predictive abilities of single-trait models ranged from 0.5 for thousand-kernel weight to 0.65 for falling number based on 10-fold cross-validations. Predictive abilities were in most cases not significantly different between single- and multi-trait models, when no phenotypic data was included for lines in the validation set. However, predictive abilities for grain protein content increased when using trait-assisted models, where the phenotypic data for Zeleny sedimentation or falling number were available for the lines in the validation set. The trait-assisted models also resulted in increased predictive abilities for Zeleny sedimentation, when phenotypic data for grain protein content was included. The latter situation could be relevant for breeding programs for improving wheat quality.

Accuracy of Single and Multi-Trait Genomic Prediction Models for Grain Yield in US Pacific Northwest Winter Wheat

Incorporating secondary correlated traits collected from high-throughput phenotyping in genomic selection (GS) models for complex traits has been demonstrated to improve accuracy. The prediction ability of different single and multiple trait partial least square (PLS) regression models for grain yield were assessed for winter wheat lines evaluated in US Pacific Northwest environments. Different populations including a diversity panel, F5, and double haploid breeding lines were evaluated in Lind and Pullman, WA between 2015 and 2018 and were genotyped with genotyping by sequencing-derived SNP markers. Prediction ability was assessed under cross-validations and independent predictions. Multi-trait covariate models were advantageous in obtaining optimal predictions for yield, especially when there is less genetic relatedness between the training and test populations. Adding multiple traits in the model improved predictions for environments with low heritability. Cross-validations resulted in the highest prediction ability (0.16) whereas independent predictions using the diversity panel to predict F5 and double haploid winter wheat breeding lines obtained the lowest (0.002). Relatedness between populations, heritability of the secondary traits, and the type of PLS model used were among the principal factors affecting prediction ability. Our results showed the relevance of using multi-trait GS models to achieve increased predictions. Genetic architecture of the target trait and genetic relatedness between populations should be taken into consideration when choosing which type of models to implement in the breeding program. An increased prediction ability for the multi-trait models indicates the potential to attain improved genetic gains for yield in wheat breeding programs through these GS approaches.

Genomic Prediction Using Multi-trait Weighted GBLUP Accounting for Heterogeneous Variances and Covariances Across the Genome.

Implicit assumption of common (co)variance for all loci in multi-trait Genomic Best Linear Unbiased Prediction (GBLUP) results in a genomic relationship matrix (G) that is common to all traits. When this assumption is violated, Bayesian whole genome regression methods may be superior to GBLUP by accounting for unequal (co)variance for all loci or genome regions. This study aimed to develop a strategy to improve the accuracy of GBLUP for multi-trait genomic prediction, using (co)variance estimates of SNP effects from Bayesian whole genome regression methods. Five generations (G1-G5, test populations) of genotype data were available by simulations based on data of 2,200 Danish Holstein cows (G0, reference population). Two correlated traits with heritabilities of 0.1 or 0.4, and a genetic correlation of 0.45 were generated. First, SNP effects and breeding values were estimated using BayesAS method, assuming (co)variance was the same for SNPs within a genome region, and different between regions. Region size was set as one SNP, 100 SNPs, a whole chromosome or whole genome. Second, posterior (co)variances of SNP effects were used to weight SNPs in construction of G matrices. In general, region size of 100 SNPs led to highest prediction accuracies using BayesAS, and wGBLUP outperformed GBLUP at this region size. Our results suggest that when genetic architectures of traits favor Bayesian methods, the accuracy of multi-trait GBLUP can be as high as the Bayesian method if SNPs are weighted by the Bayesian posterior (co)variances.

Resource allocation optimization with multi-trait genomic prediction for bread wheat (Triticum aestivum L.) baking quality

Key MessageMulti-trait genomic prediction models are useful to allocate available resources in breeding programs by targeted phenotyping of correlated traits when predicting expensive and labor-intensive quality parameters.Multi-trait genomic prediction models can be used to predict labor-intensive or expensive correlated traits where phenotyping depth of correlated traits could be larger than phenotyping depth of targeted traits, reducing resources and improving prediction accuracy. This is particularly important in the context of allocating phenotyping resource in plant breeding programs. The objective of this work was to evaluate multi-trait models predictive ability with different depth of phenotypic information from correlated traits. We evaluated 495 wheat advanced breeding lines for eight baking quality traits which were genotyped with genotyping-by-sequencing. Through different approaches for cross-validation, we evaluated the predictive ability of a single-trait model and a multi-trait model. Moreover, we evaluated different sizes of the training population (from 50 to 396 individuals) for the trait of interest, different depth of phenotypic information for correlated traits (50 and 100%) and the number of correlated traits to be used (one to three). There was no loss in the predictive ability by reducing the training population up to a 30% (149 individuals) when using correlated traits. A multi-trait model with one highly correlated trait phenotyped for both the training and testing sets was the best model considering phenotyping resources and the gain in predictive ability. The inclusion of correlated traits in the training and testing lines is a strategic approach to replace phenotyping of labor-intensive and high cost traits in a breeding program.

Genomic selection for grain yield and quality traits in durum wheat

The prediction accuracies of genomic selection depend on several factors, including the genetic architecture of target traits, the number of traits considered at a given time, and the statistical models. Here, we assessed the potential of single-trait (ST) and multi-trait (MT) genomic prediction models for durum wheat on yield and quality traits using a breeding panel (BP) of 170 varieties and advanced breeding lines, and a doubled-haploid (DH) population of 154 lines. The two populations were genotyped with the Infinium iSelect 90K SNP assay and phenotyped for various traits. Six ST-GS models (RR-BLUP, G-BLUP, BayesA, BayesB, Bayesian LASSO, and RKHS) and three MT prediction approaches (MT-BayesA, MT-Matrix, and MT-SI approaches which use economic selection index as a trait value) were applied for predicting yield, protein content, gluten index, and alveograph measures. The ST prediction accuracies ranged from 0.5 to 0.8 for the various traits and models and revealed comparable prediction accuracies for most of the traits in both populations, except BayesA and BayesB, which better predicted gluten index, tenacity, and strength in the DH population. The MT-GS models were more accurate than the ST-GS models only for grain yield in the BP. Using BP as a training set to predict the DH population resulted in poor predictions. Overall, all the six ST-GS models appear to be applicable for GS of yield and gluten strength traits in durum wheat, but we recommend the simple computational models RR-BLUP or G-BLUP for predicating single trait and MT-SI for predicting yield and protein simultaneously.

Multi-trait genomic prediction for nitrogen response indices in tropical maize hybrids

In maize breeding, genomic prediction may be an efficient tool for selecting single-crosses evaluated under abiotic stress conditions. In addition, a promising strategy is applying multiple-trait genomic prediction using selection indices (SIs), increasing genetics gains and reducing time per cycles. In this study, we aimed (i) to compare accuracy of single- and multi-trait genomic prediction (STGP; MTGP) in two maize datasets, (ii) to evaluate prediction of four selection indices that could contribute to the selection of tropical maize hybrids under contrasting nitrogen conditions, and (iii) to compare the use of linear (GBLUP) and nonlinear (RKHS/GK) kernels in STGP and MTGP analyses. For either single-trait GBLUP and RKHS analyses, the highest values obtained for accuracy were 0.40 and 0.41 using harmonic mean (HM), respectively. From multi-trait GBLUP and GK, using the combination of selection indices in MTGP seems to be suitable, increasing the accuracy. Adding grain yield and plant height in MTGP showed a slight improvement in accuracy compared to STGP. In general, there was a modest benefit of using single-trait RKHS and GK multi-trait, rather than GBLUP.

Genomic selection: the option for new robustness traits?

Genomic selection is rapidly becoming the state-of-the-art genetic selection methodology in dairy cattle breeding schemes around the world. The objective of this paper was to explore possibilities to apply genomic selection for traits related to dairy cow robustness. Deterministic simulations indicate that replacing progeny testing with genomic selection may favour genetic response for production traits at the expense of robustness traits, owing to a disproportional change in accuracies obtained across trait groups. Nevertheless, several options are available to improve the accuracy of genomic selection for robustness traits. Moreover, genomic selection opens up the opportunity to begin selection for new traits using specialised reference populations of limited size where phenotyping of large populations of animals is currently prohibitive. Reference populations for such traits may be nucleus-type herds, research herds or pooled data from (international) research experiments or research herds. The RobustMilk project has set an example for the latter approach, by collating international data for progesterone-based traits, feed intake and energy balance-related traits. Reference population design, both in terms of relatedness of the animals and variability in phenotypic performance, is important to optimise the accuracy of genomic selection. Use of indicator traits, combined with multi-trait genomic prediction models, can further contribute to improved accuracy of genomic prediction for robustness traits. Experience to date indicates that for newly recorded robustness traits that are negatively correlated with the main breeding goal, cow reference populations of ⩾10 000 are required when genotyping is based on medium- or high-density single-nucleotide polymorphism arrays. Further genotyping advances (e.g. sequencing) combined with post-genomics technologies will enhance the opportunities for (genomic) selection to improve cow robustness.