Multi-trait Genomic Prediction Research Articles

PSX-9 Comparison of single- and multiple-trait genomic predictions of cattle carcass quality traits in multibreed and purebred populations

Abstract Estimation of genetic parameters for carcass traits and their use in the evaluation of carcass quality is of considerable importance in the genetic improvement of beef cattle. The objective of this study was to compare the predictive ability of single- and multi-trait genomic prediction models for the following carcass quality traits: hot carcass weight (HCW); fat thickness (FAT); rib eye area (REA); marbling score (MARB); and peak force (PEAKF) from multibreed and purebred populations. The purebred population included 1,068 purebred Angus animals from Iowa State University whereas the multibreed population consisted of 887 animals sired by 1 of 8 purebreds (209 Angus-, 10 Brahman-, 182 Charolais-, 76 Hereford-, 44 Limousin-, 136 Maine-Anjou-, 96 Shorthorn-, and 134 South Devon- sired offspring) from the Carcass Merit Project funded by the US beef cattle industry to evaluate the genetic prediction of tenderness. After quality control of SNP markers generated with Illumina v1 50K BeadChips, 33,718 SNP markers from were used to obtain heritability and genetic correlation estimates between traits separately in either the multibreed or purebred populations. Genotyped individuals were randomly clustered into 10 groups for cross-fold validation to quantify the correlations between and predicted performance (without the individual’s data) and the realized phenotype of the individual from single- or multi-trait genomic prediction in the multibreed and purebred populations. Heritability estimates of traits from multi-trait analyses ranged from 0.42 to 0.61 in multibreed population and from 0.24 to 0.44 in the purebred population, whereas the single trait values were 0.30 to 0.53 in multibreed, and 0.23 to 0.42 in the purebred population. Estimates were greater in the multibreed population compared with the purebred population (Table 1). Compared with the single-trait analyses, multi-trait analyses resulted in higher heritability estimates for FAT, REA and PEAKF traits in the purebred population and for HCW, REA, MARB and PEAKF traits in multibreed population. Genetic correlations from multi-trait analyses indicated the genetic association between traits. In testing data sets, correlations between observed and predicted phenotypes from single-trait analyses were 4- to 7-fold greater in purebred than in the multibreed population (Table 2). These results may be due to the greater genomic relationship expected in purebred than in the multibreed population. However, the multi-trait analyses resulted in a considerable increase in correlations between observed and predicted values in multibreed population due to genetic correlations between traits.

Genomic prediction of regional-scale performance in switchgrass (Panicum virgatum) by accounting for genotype-by-environment variation and yield surrogate traits.

Switchgrass is a potential crop for bioenergy or carbon capture schemes, but further yield improvements through selective breeding are needed to encourage commercialization. To identify promising switchgrass germplasm for future breeding efforts, we conducted multisite and multitrait genomic prediction with a diversity panel of 630 genotypes from 4 switchgrass subpopulations (Gulf, Midwest, Coastal, and Texas), which were measured for spaced plant biomass yield across 10 sites. Our study focused on the use of genomic prediction to share information among traits and environments. Specifically, we evaluated the predictive ability of cross-validation (CV) schemes using only genetic data and the training set (cross-validation 1: CV1), a subset of the sites (cross-validation 2: CV2), and/or with 2 yield surrogates (flowering time and fall plant height). We found that genotype-by-environment interactions were largely due to the north-south distribution of sites. The genetic correlations between the yield surrogates and the biomass yield were generally positive (mean height r = 0.85; mean flowering time r = 0.45) and did not vary due to subpopulation or growing region (North, Middle, or South). Genomic prediction models had CV predictive abilities of -0.02 for individuals using only genetic data (CV1), but 0.55, 0.69, 0.76, 0.81, and 0.84 for individuals with biomass performance data from 1, 2, 3, 4, and 5 sites included in the training data (CV2), respectively. To simulate a resource-limited breeding program, we determined the predictive ability of models provided with the following: 1 site observation of flowering time (0.39); 1 site observation of flowering time and fall height (0.51); 1 site observation of fall height (0.52); 1 site observation of biomass (0.55); and 5 site observations of biomass yield (0.84). The ability to share information at a regional scale is very encouraging, but further research is required to accurately translate spaced plant biomass to commercial-scale sward biomass performance.

Leveraging genomics and temporal high-throughput phenotyping to enhance association mapping and yield prediction in sesame.

Sesame (Sesamum indicum) is an important oilseed crop with rising demand owing to its nutritional and health benefits. There is an urgent need to develop and integrate new genomic-based breeding strategies to meet these future demands. While genomic resources have advanced genetic research in sesame, the implementation of high-throughput phenotyping and genetic analysis of longitudinal traits remains limited. Here, we combined high-throughput phenotyping and random regression models to investigate the dynamics of plant height, leaf area index, and five spectral vegetation indices throughout the sesame growing seasons in a diversity panel. Modeling the temporal phenotypic and additive genetic trajectories revealed distinct patterns corresponding to the sesame growth cycle. We also conducted longitudinal genomic prediction and association mapping of plant height using various models and cross-validation schemes. Moderate prediction accuracy was obtained when predicting new genotypes at each time point, and moderate to high values were obtained when forecasting future phenotypes. Association mapping revealed three genomic regions in linkage groups 6, 8, and 11, conferring trait variation over time and growth rate. Furthermore, we leveraged correlations between the temporal trait and seed-yield and applied multi-trait genomic prediction. We obtained an improvement over single-trait analysis, especially when phenotypes from earlier time points were used, highlighting the potential of using a high-throughput phenotyping platform as a selection tool. Our results shed light on the genetic control of longitudinal traits in sesame and underscore the potential of high-throughput phenotyping to detect a wide range of traits and genotypes that can inform sesame breeding efforts to enhance yield.

Enhancing the potential of phenomic and genomic prediction in winter wheat breeding using high-throughput phenotyping and deep learning.

Integrating high-throughput phenotyping (HTP) based traits into phenomic and genomic selection (GS) can accelerate the breeding of high-yielding and climate-resilient wheat cultivars. In this study, we explored the applicability of Unmanned Aerial Vehicles (UAV)-assisted HTP combined with deep learning (DL) for the phenomic or multi-trait (MT) genomic prediction of grain yield (GY), test weight (TW), and grain protein content (GPC) in winter wheat. Significant correlations were observed between agronomic traits and HTP-based traits across different growth stages of winter wheat. Using a deep neural network (DNN) model, HTP-based phenomic predictions showed robust prediction accuracies for GY, TW, and GPC for a single location with R2 of 0.71, 0.62, and 0.49, respectively. Further prediction accuracies increased (R2 of 0.76, 0.64, and 0.75) for GY, TW, and GPC, respectively when advanced breeding lines from multi-locations were used in the DNN model. Prediction accuracies for GY varied across growth stages, with the highest accuracy at the Feekes 11 (Milky ripe) stage. Furthermore, forward prediction of GY in preliminary breeding lines using DNN trained on multi-location data from advanced breeding lines improved the prediction accuracy by 32% compared to single-location data. Next, we evaluated the potential of incorporating HTP-based traits in multi-trait genomic selection (MT-GS) models in the prediction of GY, TW, and GPC. MT-GS, models including UAV data-based anthocyanin reflectance index (ARI), green chlorophyll index (GCI), and ratio vegetation index 2 (RVI_2) as covariates demonstrated higher predictive ability (0.40, 0.40, and 0.37, respectively) as compared to single-trait model (0.23) for GY. Overall, this study demonstrates the potential of integrating HTP traits into DL-based phenomic or MT-GS models for enhancing breeding efficiency.

Single- and multi-trait genomic prediction and genome-wide association analysis of grain yield and micronutrient-related traits in ICARDA wheat under drought environment.

Globally, over 2 billion people suffer from malnutrition due to inadequate intake of micronutrients. Genomic-assisted breeding is identified as a valuable method to facilitate developing new improved plant varieties targeting grain yield and micronutrient-related traits. In this study, a genome-wide association study (GWAS) and single- and multi-trait-based genomic prediction (GP) analysis was conducted using a set of 252 elite wheat genotypes from the International Center for Agricultural Research in Dry Areas (ICARDA). The objective was to identify linked SNP markers, putative candidate genes and to evaluate the genomic estimated breeding values (GEBVs) of grain yield and micronutrient-related traits.. For this purpose, a field trial was conducted at a drought-prone station, Merchouch, Morocco for 2 consecutive years (2018 and 2019) followed by GWAS and genomic prediction analysis with 10,173 quality SNP markers. The studied genotypes exhibited a significant genotypic variation in grain yield and micronutrient-related traits. The GWAS analysis identified highly significantly associated markers and linked putative genes on chromosomes 1B and 2B for zinc (Zn) and iron (Fe) contents, respectively. The genomic predictive ability of selenium (Se) and Fe traits with the multi-trait-based GP GBLUP model was 0.161 and 0.259 improving by 6.62 and 4.44%, respectively, compared to the corresponding single-trait-based models. The identified significantly linked SNP markers, associated putative genes, and developed GP models could potentially facilitate breeding programs targeting to improve the overall genetic gain of wheat breeding for grain yield and biofortification of micronutrients via marker-assisted (MAS) and genomic selection (GS) methods.

Multi-Trait Single-Step Genomic Prediction for Milk Yield and Milk Components for Polish Holstein Population.

The objective of our study was to evaluate the predictive ability of a multi-trait genomic prediction model that accounts for interactions between marker effects to estimate heritability and genetic correlations of traits including 305-day milk yield, milk fat percentage, milk protein percentage, milk lactose percentage, and milk dry matter percentage in the Polish Holstein Friesian cow population. For this aim, 14,742 SNP genotype records for 586 Polish Holstein Friesian dairy cows from Poland were used. Single-Trait-ssGBLUP (ST) and Multi-Trait-ssGBLUP (MT) methods were used for estimation. We examined 305-day milk yield (MY, kg), milk fat percentage (MF, %), milk protein percentage (MP, %), milk lactose percentage (ML, %), and milk dry matter percentage (MDM, %). The results showed that the highest marker effect rank correlation was found between milk fat percentage and milk dry matter. The weakest marker effect rank correlation was found between ML and all other traits. Obtained accuracies of this study were between 0.770 and 0.882, and 0.773 and 0.876 for MT and ST, respectively, which were acceptable values. All estimated bias values were positive, which is proof of underestimation. The highest heritability value was obtained for MP (0.3029) and the lowest heritability value was calculated for ML (0.2171). Estimated heritability values were low for milk yield and milk composition as expected. The strongest genetic correlation was estimated between MDM and MF (0.4990) and the weakest genetic correlation was estimated between MY and ML (0.001). The genetic relations with milk yield were negative and can be ignored as they were not significant. In conclusion, multi-trait genomic prediction can be more beneficial than single-trait genomic prediction.

Partial least squares enhance multi-trait genomic prediction of potato cultivars in new environments

It is of paramount importance in plant breeding to have methods dealing with large numbers of predictor variables and few sample observations, as well as efficient methods for dealing with high correlation in predictors and measured traits. This paper explores in terms of prediction performance the partial least squares (PLS) method under single-trait (ST) and multi-trait (MT) prediction of potato traits. The first prediction was for tested lines in tested environments under a five-fold cross-validation (5FCV) strategy and the second prediction was for tested lines in untested environments (herein denoted as leave one environment out cross validation, LOEO). There was a good performance in terms of predictions (with accuracy mostly > 0.5 for Pearson’s correlation) the accuracy of 5FCV was better than LOEO. Hence, we have empirical evidence that the ST and MT PLS framework is a very valuable tool for prediction in the context of potato breeding data.

(Quasi) multitask support vector regression with heuristic hyperparameter optimization for whole-genome prediction of complex traits: a case study with carcass traits in broilers.

This study investigates nonlinear kernels for multitrait (MT) genomic prediction using support vector regression (SVR) models. We assessed the predictive ability delivered by single-trait (ST) and MT models for 2 carcass traits (CT1 and CT2) measured in purebred broiler chickens. The MT models also included information on indicator traits measured in vivo [Growth and feed efficiency trait (FE)]. We proposed an approach termed (quasi) multitask SVR (QMTSVR), with hyperparameter optimization performed via genetic algorithm. ST and MT Bayesian shrinkage and variable selection models [genomic best linear unbiased predictor (GBLUP), BayesC (BC), and reproducing kernel Hilbert space (RKHS) regression] were employed as benchmarks. MT models were trained using 2 validation designs (CV1 and CV2), which differ if the information on secondary traits is available in the testing set. Models' predictive ability was assessed with prediction accuracy (ACC; i.e. the correlation between predicted and observed values, divided by the square root of phenotype accuracy), standardized root-mean-squared error (RMSE*), and inflation factor (b). To account for potential bias in CV2-style predictions, we also computed a parametric estimate of accuracy (ACCpar). Predictive ability metrics varied according to trait, model, and validation design (CV1 or CV2), ranging from 0.71 to 0.84 for ACC, 0.78 to 0.92 for RMSE*, and between 0.82 and 1.34 for b. The highest ACC and smallest RMSE* were achieved with QMTSVR-CV2 in both traits. We observed that for CT1, model/validation design selection was sensitive to the choice of accuracy metric (ACC or ACCpar). Nonetheless, the higher predictive accuracy of QMTSVR over MTGBLUP and MTBC was replicated across accuracy metrics, besides the similar performance between the proposed method and the MTRKHS model. Results showed that the proposed approach is competitive with conventional MT Bayesian regression models using either Gaussian or spike-slab multivariate priors.

Improved multi-trait prediction of wheat end-product quality traits by integrating NIR-predicted phenotypes.

Historically, end-product quality testing has been costly and required large flour samples; therefore, it was generally implemented in the late phases of variety development, imposing a huge cost on the breeding effort and effectiveness. High genetic correlations of end-product quality traits with higher throughput and nondestructive testing technologies, such as near-infrared (NIR), could enable early-stage testing and effective selection of these highly valuable traits in a multi-trait genomic prediction model. We studied the impact on prediction accuracy in genomic best linear unbiased prediction (GBLUP) of adding NIR-predicted secondary traits for six end-product quality traits (crumb yellowness, water absorption, texture hardness, flour yield, grain protein, flour swelling volume). Bread wheat lines (1,400-1,900) were measured across 8 years (2012-2019) for six end-product quality traits with standard laboratory assays and with NIR, which were combined to generate predicted data for approximately 27,000 lines. All lines were genotyped with the Infinium™ Wheat Barley 40K BeadChip and imputed using exome sequence data. End-product and NIR phenotypes were genetically correlated (0.5-0.83, except for flour swelling volume 0.19). Prediction accuracies of end-product traits ranged between 0.28 and 0.64 and increased by 30% through the inclusion of NIR-predicted data compared to single-trait analysis. There was a high correlation between the multi-trait prediction accuracy and genetic correlations between end-product and NIR-predicted data (0.69-0.77). Our forward prediction validation revealed a gradual increase in prediction accuracy when adding more years to the multi-trait model. Overall, we achieved genomic prediction accuracy at a level that enables selection for end-product quality traits early in the breeding cycle.

Multi-trait genomic selection improves the prediction accuracy of end-use quality traits in hard winter wheat.

Improvement of end-use quality remains one of the most important goals in hard winter wheat (HWW) breeding. Nevertheless, the evaluation of end-use quality traits is confined to later development generations owing to resource-intensive phenotyping. Genomic selection (GS) has shown promise in facilitating selection for end-use quality; however, lower prediction accuracy (PA) for complex traits remains a challenge in GS implementation. Multi-trait genomic prediction (MTGP) models can improve PA for complex traits by incorporating information on correlated secondary traits, but these models remain to be optimized in HWW. A set of advanced breeding lines from 2015 to 2021 were genotyped with 8725 single-nucleotide polymorphisms and was used to evaluate MTGP to predict various end-use quality traits that are otherwise difficult to phenotype in earlier generations. The MTGP model outperformed the ST model with up to a twofold increase in PA. For instance, PA was improved from 0.38 to 0.75 for bake absorption and from 0.32 to 0.52 for loaf volume. Further, we compared MTGP models by including different combinations of easy-to-score traits as covariates to predict end-use quality traits. Incorporation of simple traits, such as flour protein (FLRPRO) and sedimentation weight value (FLRSDS), substantially improved the PA of MT models. Thus, the rapid low-cost measurement of traits like FLRPRO and FLRSDS can facilitate the use of GP to predict mixograph and baking traits in earlier generations and provide breeders an opportunity for selection on end-use quality traits by culling inferior lines to increase selection accuracy and genetic gains.

Genetic parameters and multi-trait genomic prediction for hemoparasites infection levels in cattle

Babesiosis and anaplasmosis are tick-borne diseases that substantially affect the economic outcomes of livestock production in tropical countries. This study aimed to evaluate the genetics of resistance to infection caused by these parasites through the estimation of heritabilities and genetic correlations of infection levels among Babesia bigemina (BigIL), B. bovis (BovIL), and Anaplasma marginale (AmIL), and tick counts (TC). The predictive ability of single and multi-trait genomic prediction models was evaluated through various combinations of these traits. To our knowledge, this is the first genomic study to examine BigIL and AmIL. Infection levels of BigIL (n = 1,882), BovIL (n = 1,858), and AmIL (n = 1,523) were estimated from blood samples using real-time PCR. TC phenotypes (n = 5,867) were obtained by counting the number of parasites larger than 4.5 mm from the right-hand side of each animal. Genotypic data were available for 3,977 animals which were then imputed up to ∼777,000 SNP and, after quality control, 502,398 SNP remained for downstream analyses. Variance components for BigIL and AmIL and the genetic correlations between traits were estimated using a Bayesian approach. The single-step best linear unbiased prediction was used to estimate genomic breeding values (GBV). The heritability estimates for BigIL and AmIL were low at 0.094 and 0.090, respectively, suggesting high environmental influence levels for both traits. The genetic correlations between tick count and infection levels for BigIL (0.239), BovIL (0.160), and AmIL (-0.019) were low, as well as the correlation between AmIL and BovIL (0.043). The genetic correlations between BigIL and BovIL (0.524) and BigIL and AmIL (0.793) were high, which contributed to improved GBV accuracies when these traits were combined in multi-trait models in comparison to single-trait models. These results suggested that multi-trait genomic prediction models of infection levels for tick-borne diseases are preferable to single-trait models. Additionally, our results indicated that the TC data and the GBV based on them are not useful for predicting infection levels of BigIL, BovIL, and AmIL.

Single and multi-trait genomic prediction for agronomic traits in Euterpe edulis.

Popularly known as juçaizeiro, Euterpe edulis has been gaining prominence in the fruit growing sector and has demanded the development of superior genetic materials. Since it is a native species and still little studied, the application of more sophisticated techniques can result in higher gains with less time. Until now, there are no studies that apply genomic prediction for this crop, especially in multi-trait analysis. In this sense, this study aimed to apply new methods and breeding techniques for the juçaizeiro, to optimize this breeding program through the application of genomic prediction. This data consisted of 275 juçaizeiro genotypes from a population of Rio Novo do Sul-ES, Brazil. The genomic prediction was performed using the multi-trait (G-BLUP MT) and single-trait (G-BLUP ST) models and the selection of superior genotypes was based on a selection index. Similar results for predictive ability were observed for both models. However, the G-BLUP ST model provided greater selection gains when compared to the G-BLUP MT. For this reason, the genomic estimated breeding values (GEBVs) from the G-BLUP ST, were used to select the six superior genotypes (UFES.A.RN.390, UFES.A.RN.386, UFES.A.RN.080, UFES.A.RN.383, UFES.S.RN.098, and UFES.S.RN.093). This was intended to provide superior genetic materials for the development of seedlings and implantation of productive orchards, which will meet the demands of the productive, industrial and consumer market.

Multi-trait genomic prediction in pigs using single and multistep methods based on the absorption of ungenotyped animals.

Many quantitative traits measured in breeding programs are genetically correlated. The genetic correlations between the traits indicate that the measurement of one trait carries information on others. To benefit from this information, multi-trait genomic prediction (MTGP) is preferable to use. However, MTGP is more difficult to implement compared to single-trait genomic prediction (STGP), and even more challenging for the goal to exploit not only the information on other traits but also the information on ungenotyped animals. This could be accomplished using both single and multistep methods. The single-step method was achieved by implementing a single-step genomic best linear unbiased prediction (ssGBLUP) approach using a multi-trait model. Here, we examined a multistep analysis based on an approach called "Absorption" to achieve this goal. The Absorption approach absorbed all available information including the phenotypic information on ungenotyped animals as well as the information on other traits if applicable, into mixed model equations of genotyped animals. The multistep analysis included (1) to apply the Absorption approach that exploits all available information and (2) to implement genomic BLUP (GBLUP) prediction on the absorbed dataset. In this study, the ssGBLUP and multistep analysis were applied to 5 traits in Duroc pigs, which were slaughter percentage, feed consumption from 40 to 120 kg (FC40_120), days of growth from 40 to 120 kg (D40_120), age at 40 kg (A40) and lean meat percentage. The results showed that MTGP yielded higher accuracy than STGP, which on average was 0.057 higher for the multistep method and 0.045 higher for ssGBLUP. The multistep method achieved similar prediction accuracy as ssGBLUP. However, the prediction bias of the multistep method was in general lower than that of ssGBLUP.

Evaluating metabolic and genomic data for predicting grain traits under high night temperature stress in rice.

The asymmetric increase in average nighttime temperatures relative to increase in average daytime temperatures due to climate change is decreasing grain yield and quality in rice. Therefore, a better genome-level understanding of the impact of higher night temperature stress on the weight of individual grains is essential for future development of more resilient rice. We investigated the utility of metabolites obtained from grains to classify high night temperature (HNT) conditions of genotypes, and metabolites and single-nucleotide polymorphisms (SNPs) to predict grain length, width, and perimeter phenotypes using a rice diversity panel. We found that the metabolic profiles of rice genotypes alone could be used to classify control and HNT conditions with high accuracy using random forest or extreme gradient boosting. Best linear unbiased prediction and BayesC showed greater metabolic prediction performance than machine learning models for grain-size phenotypes. Metabolic prediction was most effective for grain width, resulting in the highest prediction performance. Genomic prediction performed better than metabolic prediction. Integrating metabolites and genomics simultaneously in a prediction model slightly improved prediction performance. We did not observe a difference in prediction between the control and HNT conditions. Several metabolites were identified as auxiliary phenotypes that could be used to enhance the multi-trait genomic prediction of grain-size phenotypes. Our results showed that, in addition to SNPs, metabolites collected from grains offer rich information to perform predictive analyses, including classification modeling of HNT responses and regression modeling of grain-size-related phenotypes in rice.

Sparse multi-trait genomic prediction under balanced incomplete block design.

Sparse testing is essential to increase the efficiency of the genomic selection methodology, as the same efficiency (in this case prediction power) can be obtained while using less genotypes evaluated in the fields. For this reason, it is important to evaluate the existing methods for performing the allocation of lines to environments. With this goal, four methods (M1-M4) to allocate lines to environments were evaluated under the context of a multi-trait genomic prediction problem: M1 denotes the allocation of a fraction (subset) of lines in all locations, M2 denotes the allocation of a fraction of lines with some shared lines in locations but not arranged based on the balanced incomplete block design (BIBD) principle, M3 denotes the random allocation of a subset of lines to locations, and M4 denotes the allocation of a subset of lines to locations using the BIBD principle. The evaluation was done using seven real multi-environment data sets common in plant breeding programs. We found that the best method was M4 and the worst was M1, while no important differences were found between M3 and M4. We concluded that M4 and M3 are efficient in the context of sparse testing for multi-trait prediction.

Multi-Trait Genomic Prediction Improves Accuracy of Selection among Doubled Haploid Lines in Maize

Recent advances in maize doubled haploid (DH) technology have enabled the development of large numbers of DH lines quickly and efficiently. However, testing all possible hybrid crosses among DH lines is a challenge. Phenotyping haploid progenitors created during the DH process could accelerate the selection of DH lines. Based on phenotypic and genotypic data of a DH population and its corresponding haploids, we compared phenotypes and estimated genetic correlations between the two populations, compared genomic prediction accuracy of multi-trait models against conventional univariate models within the DH population, and evaluated whether incorporating phenotypic data from haploid lines into a multi-trait model could better predict performance of DH lines. We found significant phenotypic differences between DH and haploid lines for nearly all traits; however, their genetic correlations between populations were moderate to strong. Furthermore, a multi-trait model taking into account genetic correlations between traits in the single-environment trial or genetic covariances in multi-environment trials can significantly increase genomic prediction accuracy. However, integrating information of haploid lines did not further improve our prediction. Our findings highlight the superiority of multi-trait models in predicting performance of DH lines in maize breeding, but do not support the routine phenotyping and selection on haploid progenitors of DH lines.

Multi-trait genomic prediction improves selection accuracy for enhancing seed mineral concentrations in pea.

Multi-trait genomic selection (MT-GS) has the potential to improve predictive ability by maximizing the use of information across related genotypes and genetically correlated traits. In this study, we extended the use of sparse phenotyping method into the MT-GS framework by split testing of entries to maximize borrowing of information across genotypes and predict missing phenotypes for targeted traits without additional phenotyping expenditure. Using 300 advanced breeding lines from North Dakota State University (NDSU) pulse breeding program and ∼200 USDA accessions that were evaluated for 10 nutritional traits, our results show that the proposed sparse phenotyping aided MT-GS can further improve predictive ability by >12% across traits compared with univariate (UNI) genomic selection. The proposed strategy departed from the previous reports that weak genetic correlation is a limitation to the advantage of MT-GS over UNI genomic selection, which was evident in the partially balanced phenotyping-enabled MT-GS. Our results point to heritability and genetic correlation between traits as possible metrics to optimize and further improve the estimation of model parameters, and ultimately, prediction performance. Overall, our study offers a new approach to optimize the prediction performance using the MT-GS and further highlight strategy to maximize the efficiency of GS in a plant breeding program. The sparse-testing-aided MT-GS proposed in this study can be further extended to multi-environment, multi-trait GS to improve prediction performance and further reduce the cost of phenotyping and time-consuming data collection process.

Time-series multispectral imaging in soybean for improving biomass and genomic prediction accuracy.

Multispectral (MS) imaging enables the measurement of characteristics important for increasing the prediction accuracy of genotypic and phenotypic values for yield-related traits. In this study, we evaluated the potential application of temporal MS imaging for the prediction of aboveground biomass (AGB) in soybean [Glycine max (L.) Merr.]. Field experiments with 198 accessions of soybean were conducted with four different irrigation levels. Five vegetation indices (VIs) were calculated using MS images from soybean canopies from early vegetative to early reproductive stage. To predict the genotypic values of AGB, VIs at the different growth stages were used as secondary traits in a multitrait genomic prediction. The prediction accuracy of the genotypic values of AGB from MS and genomic data largely outperformed that of the genomic data alone before the flowering stage (90% of accessions did not flower), suggesting that it would be possible to determine cross-combinations based on the predicted genotypic values of AGB. We compared the prediction accuracy of a model using the five VIs and a model using only one VI to predict the phenotypic values of AGB and found that the difference in prediction accuracy decreased over time at all irrigation levels except for the most severe drought. The difference in the most severe drought was not as small as that in the other treatments. Only the prediction accuracy of a model using the five VIs in the most severe droughts gradually increased over time. Therefore, the optimal timing for MS imaging may depend on the irrigation levels.

A Comparison of Three Machine Learning Methods for Multivariate Genomic Prediction Using the Sparse Kernels Method (SKM) Library.

Genomic selection (GS) changed the way plant breeders select genotypes. GS takes advantage of phenotypic and genotypic information to training a statistical machine learning model, which is used to predict phenotypic (or breeding) values of new lines for which only genotypic information is available. Therefore, many statistical machine learning methods have been proposed for this task. Multi-trait (MT) genomic prediction models take advantage of correlated traits to improve prediction accuracy. Therefore, some multivariate statistical machine learning methods are popular for GS. In this paper, we compare the prediction performance of three MT methods: the MT genomic best linear unbiased predictor (GBLUP), the MT partial least squares (PLS) and the multi-trait random forest (RF) methods. Benchmarking was performed with six real datasets. We found that the three investigated methods produce similar results, but under predictors with genotype (G) and environment (E), that is, E + G, the MT GBLUP achieved superior performance, whereas under predictors E + G + genotype environment (GE) and G + GE, random forest achieved the best results. We also found that the best predictions were achieved under the predictors E + G and E + G + GE. Here, we also provide the R code for the implementation of these three statistical machine learning methods in the sparse kernel method (SKM) library, which offers not only options for single-trait prediction with various statistical machine learning methods but also some options for MT predictions that can help to capture improved complex patterns in datasets that are common in genomic selection.

Genomic Prediction Strategies for Dry-Down-Related Traits in Maize.

For efficient mechanical harvesting, low grain moisture content at harvest time is essential. Dry-down rate (DR), which refers to the reduction in grain moisture content after the plants enter physiological maturity, is one of the main factors affecting the amount of moisture in the kernels. Dry-down rate is estimated using kernel moisture content at physiological maturity and at harvest time; however, measuring kernel water content at physiological maturity, which is sometimes referred as kernel water content at black layer formation (BWC), is time-consuming and resource-demanding. Therefore, inferring BWC from other correlated and easier to measure traits could improve the efficiency of breeding efforts for dry-down-related traits. In this study, multi-trait genomic prediction models were used to estimate genetic correlations between BWC and water content at harvest time (HWC) and flowering time (FT). The results show there is moderate-to-high genetic correlation between the traits (0.24–0.66), which supports the use of multi-trait genomic prediction models. To investigate genomic prediction strategies, several cross-validation scenarios representing possible implementations of genomic prediction were evaluated. The results indicate that, in most scenarios, the use of multi-trait genomic prediction models substantially increases prediction accuracy. Furthermore, the inclusion of historical records for correlated traits can improve prediction accuracy, even when the target trait is not measured on all the plots in the training set.