Use of multiple traits genomic prediction, genotype by environment interactions and spatial effect to improve prediction accuracy in yield data.

Genomic selection has been extensively implemented in plant breeding schemes. Genomic selection incorporates dense genome-wide markers to predict the breeding values for important traits based on information from genotype and phenotype records on traits of interest in a reference population. To date, most relevant investigations have been performed using single trait genomic prediction models (STGP). However, records for several traits at once are usually documented for breeding lines in commercial breeding programs. By incorporating benefits from genetic characterizations of correlated phenotypes, multiple trait genomic prediction (MTGP) may be a useful tool for improving prediction accuracy in genetic evaluations. The objective of this study was to test whether the use of MTGP and including proper modeling of spatial effects can improve the prediction accuracy of breeding values in commercial barley and wheat breeding lines. We genotyped 1,317 spring barley and 1,325 winter wheat lines from a commercial breeding program with the Illumina 9K barley and 15K wheat SNP-chip (respectively) and phenotyped them across multiple years and locations. Results showed that the MTGP approach increased correlations between future performance and estimated breeding value of yields by 7% in barley and by 57% in wheat relative to using the STGP approach for each trait individually. Analyses combining genomic data, pedigree information, and proper modeling of spatial effects further increased the prediction accuracy by 4% in barley and 3% in wheat relative to the model using genomic relationships only. The prediction accuracy for yield in wheat and barley yield trait breeding, were improved by combining MTGP and spatial effects in the model.

Genomic Prediction in Tetraploid Ryegrass Using Allele Frequencies Based on Genotyping by Sequencing.

Perennial ryegrass is an outbreeding forage species and is one of the most widely used forage grasses in temperate regions. The aim of this study was to investigate the possibility of implementing genomic prediction in tetraploid perennial ryegrass, to study the effects of different sequencing depth when using genotyping-by-sequencing (GBS), and to determine optimal number of single-nucleotide polymorphism (SNP) markers and sequencing depth for GBS data when applied in tetraploids. A total of 1,515 F2 tetraploid ryegrass families were included in the study and phenotypes and genotypes were scored on family-pools. The traits considered were dry matter yield (DM), rust resistance (RUST), and heading date (HD). The genomic information was obtained in the form of allele frequencies of pooled family samples using GBS. Different SNP filtering strategies were designed. The strategies included filtering out SNPs having low average depth (FILTLOW), having high average depth (FILTHIGH), and having both low average and high average depth (FILTBOTH). In addition, SNPs were kept randomly with different data sizes (RAN). The accuracy of genomic prediction was evaluated by using a "leave single F2 family out" cross validation scheme, and the predictive ability and bias were assessed by correlating phenotypes corrected for fixed effects with predicted additive breeding values. Among all the filtering scenarios, the highest estimates for genomic heritability of family means were 0.45, 0.74, and 0.73 for DM, HD and RUST, respectively. The predictive ability generally increased as the number of SNPs included in the analysis increased. The highest predictive ability for DM was 0.34 (137,191 SNPs having average depth higher than 10), for HD was 0.77 (185,297 SNPs having average depth lower than 60), and for RUST was 0.55 (188,832 SNPs having average depth higher than 1). Genomic prediction can help to optimize the breeding of tetraploid ryegrass. GBS data including about 80-100 K SNPs are needed for accurate prediction of additive breeding values in tetraploid ryegrass. Using only SNPs with sequencing depth between 10 and 20 gave highest predictive ability, and showed the potential to obtain accurate prediction from medium-low coverage GBS in tetraploids.

Optimized Use of Low-Depth Genotyping-by-Sequencing for Genomic Prediction Among Multi-Parental Family Pools and Single Plants in Perennial Ryegrass (Lolium perenne L.).

Ryegrass single plants, bi-parental family pools, and multi-parental family pools are often genotyped, based on allele-frequencies using genotyping-by-sequencing (GBS) assays. GBS assays can be performed at low-coverage depth to reduce costs. However, reducing the coverage depth leads to a higher proportion of missing data, and leads to a reduction in accuracy when identifying the allele-frequency at each locus. As a consequence of the latter, genomic relationship matrices (GRMs) will be biased. This bias in GRMs affects variance estimates and the accuracy of GBLUP for genomic prediction (GBLUP-GP). We derived equations that describe the bias from low-coverage sequencing as an effect of binomial sampling of sequence reads, and allowed for any ploidy level of the sample considered. This allowed us to combine individual and pool genotypes in one GRM, treating pool-genotypes as a polyploid genotype, equal to the total ploidy-level of the parents of the pool. Using simulated data, we verified the magnitude of the GRM bias at different coverage depths for three different kinds of ryegrass breeding material: individual genotypes from single plants, pool-genotypes from F2 families, and pool-genotypes from synthetic varieties. To better handle missing data, we also tested imputation procedures, which are suited for analyzing allele-frequency genomic data. The relative advantages of the bias-correction and the imputation of missing data were evaluated using real data. We examined a large dataset, including single plants, F2 families, and synthetic varieties genotyped in three GBS assays, each with a different coverage depth, and evaluated them for heading date, crown rust resistance, and seed yield. Cross validations were used to test the accuracy using GBLUP approaches, demonstrating the feasibility of predicting among different breeding material. Bias-corrected GRMs proved to increase predictive accuracies when compared with standard approaches to construct GRMs. Among the imputation methods we tested, the random forest method yielded the highest predictive accuracy. The combinations of these two methods resulted in a meaningful increase of predictive ability (up to 0.09). The possibility of predicting across individuals and pools provides new opportunities for improving ryegrass breeding schemes.

Genome-Wide Association Studies and Comparison of Models and Cross-Validation Strategies for Genomic Prediction of Quality Traits in Advanced Winter Wheat Breeding Lines.

The aim of the this study was to identify SNP markers associated with five important wheat quality traits (grain protein content, Zeleny sedimentation, test weight, thousand-kernel weight, and falling number), and to investigate the predictive abilities of GBLUP and Bayesian Power Lasso models for genomic prediction of these traits. In total, 635 winter wheat lines from two breeding cycles in the Danish plant breeding company Nordic Seed A/S were phenotyped for the quality traits and genotyped for 10,802 SNPs. GWAS were performed using single marker regression and Bayesian Power Lasso models. SNPs with large effects on Zeleny sedimentation were found on chromosome 1B, 1D, and 5D. However, GWAS failed to identify single SNPs with significant effects on the other traits, indicating that these traits were controlled by many QTL with small effects. The predictive abilities of the models for genomic prediction were studied using different cross-validation strategies. Leave-One-Out cross-validations resulted in correlations between observed phenotypes corrected for fixed effects and genomic estimated breeding values of 0.50 for grain protein content, 0.66 for thousand-kernel weight, 0.70 for falling number, 0.71 for test weight, and 0.79 for Zeleny sedimentation. Alternative cross-validations showed that the genetic relationship between lines in training and validation sets had a bigger impact on predictive abilities than the number of lines included in the training set. Using Bayesian Power Lasso instead of GBLUP models, gave similar or slightly higher predictive abilities. Genomic prediction based on all SNPs was more effective than prediction based on few associated SNPs.

ADAM-Plant: A Software for Stochastic Simulations of Plant Breeding From Molecular to Phenotypic Level and From Simple Selection to Complex Speed Breeding Programs.

Making decisions on plant breeding programs require plant breeders to be able to test different breeding strategies by taking into account all the crucial factors affecting crop genetic improvement. Due to the complexity of the decisions, computer simulation serves as an important tool for researchers and plant breeders. This paper describes ADAM-plant, which is a computer software that models breeding schemes for self-pollinated and cross-pollinated crop plants using stochastic simulation. The program simulates a population of plants and traces the genetic changes in the population under different breeding scenarios. It takes into account different population structures, genomic models, selection (strategies and units) and crossing strategies. It also covers important features e.g., allowing users to perform genomic selection (GS) and speed breeding, simulate genotype-by-environment interactions using multiple trait approach, simulate parallel breeding cycles and consider plot sizes. In addition, the software can be used to simulate datasets produced from very complex breeding program in order to test new statistical methodology to analyze such data. As an example, three wheat-breeding strategies were simulated in the current study: (1) phenotypic selection, (2) GS, and (3) speed breeding with genomic information. The results indicate that the genetic gain can be doubled by GS compared to phenotypic selection and genetic gain can be further increased considerably by speed breeding. In conclusion, ADAM-plant is an important tool for comparing strategies for plant breeding and for estimating the effects of allocation of different resources to the breeding program. In the current study, it was used to compare different methodologies for utilizing genomic information in cereal breeding programs for selection of best-fit breeding strategy as per available resources.

Optimizing Training Population Size and Genotyping Strategy for Genomic Prediction Using Association Study Results and Pedigree Information. A Case of Study in Advanced Wheat Breeding Lines.

Wheat breeding programs generate a large amount of variation which cannot be completely explored because of limited phenotyping throughput. Genomic prediction (GP) has been proposed as a new tool which provides breeding values estimations without the need of phenotyping all the material produced but only a subset of it named training population (TP). However, genotyping of all the accessions under analysis is needed and, therefore, optimizing TP dimension and genotyping strategy is pivotal to implement GP in commercial breeding schemes. Here, we explored the optimum TP size and we integrated pedigree records and genome wide association studies (GWAS) results to optimize the genotyping strategy. A total of 988 advanced wheat breeding lines were genotyped with the Illumina 15K SNPs wheat chip and phenotyped across several years and locations for yield, lodging, and starch content. Cross-validation using the largest possible TP size and all the SNPs available after editing (~11k), yielded predictive abilities (rGP) ranging between 0.5-0.6. In order to explore the Training population size, rGP were computed using progressively smaller TP. These exercises showed that TP of around 700 lines were enough to yield the highest observed rGP. Moreover, rGP were calculated by randomly reducing the SNPs number. This showed that around 1K markers were enough to reach the highest observed rGP. GWAS was used to identify markers associated with the traits analyzed. A GWAS-based selection of SNPs resulted in increased rGP when compared with random selection and few hundreds SNPs were sufficient to obtain the highest observed rGP. For each of these scenarios, advantages of adding the pedigree information were shown. Our results indicate that moderate TP sizes were enough to yield high rGP and that pedigree information and GWAS results can be used to greatly optimize the genotyping strategy.

Genomic Prediction of Seed Quality Traits Using Advanced Barley Breeding Lines.

Genomic selection was recently introduced in plant breeding. The objective of this study was to develop genomic prediction for important seed quality parameters in spring barley. The aim was to predict breeding values without expensive phenotyping of large sets of lines. A total number of 309 advanced spring barley lines tested at two locations each with three replicates were phenotyped and each line was genotyped by Illumina iSelect 9Kbarley chip. The population originated from two different breeding sets, which were phenotyped in two different years. Phenotypic measurements considered were: seed size, protein content, protein yield, test weight and ergosterol content. A leave-one-out cross-validation strategy revealed high prediction accuracies ranging between 0.40 and 0.83. Prediction across breeding sets resulted in reduced accuracies compared to the leave-one-out strategy. Furthermore, predicting across full and half-sib-families resulted in reduced prediction accuracies. Additionally, predictions were performed using reduced marker sets and reduced training population sets. In conclusion, using less than 200 lines in the training set can result in low prediction accuracy, and the accuracy will then be highly dependent on the family structure of the selected training set. However, the results also indicate that relatively small training sets (200 lines) are sufficient for genomic prediction in commercial barley breeding. In addition, our results indicate a minimum marker set of 1,000 to decrease the risk of low prediction accuracy for some traits or some families.

Genomic dissection and prediction of heading date in perennial ryegrass.

BACKGROUND: Genomic selection (GS) has become a commonly used technology in animal breeding. In crops, it is expected to significantly improve the genetic gains per unit of time. So far, its implementation in plant breeding has been mainly investigated in species farmed as homogeneous varieties. Concerning crops farmed in family pools, only a few theoretical studies are currently available. Here, we test the opportunity to implement GS in breeding of perennial ryegrass, using real data from a forage breeding program. Heading date was chosen as a model trait, due to its high heritability and ease of assessment. Genome Wide Association analysis was performed to uncover the genetic architecture of the trait. Then, Genomic Prediction (GP) models were tested and prediction accuracy was compared to the one obtained in traditional Marker Assisted Selection (MAS) methods. RESULTS: Several markers were significantly associated with heading date, some locating within or proximal to genes with a well-established role in floral regulation. GP models gave very high accuracies, which were significantly better than those obtained through traditional MAS. Accuracies were higher when predictions were made from related families and from larger training populations, whereas predicting from unrelated families caused the variance of the estimated breeding values to be biased downwards. CONCLUSIONS: We have demonstrated that there are good perspectives for GS implementation in perennial ryegrass breeding, and that problems resulting from low linkage disequilibrium (LD) can be reduced by the presence of structure and related families in the breeding population. While comprehensive Genome Wide Association analysis is difficult in species with extremely low LD, we did identify variants proximal to genes with a known role in flowering time (e.g. CONSTANS and Phytochrome C).

Association Mapping for Fruit, Plant and Leaf Morphology Traits in Eggplant.

An eggplant (Solanum melongena) association panel of 191 accessions, comprising a mixture of breeding lines, old varieties and landrace selections was SNP genotyped and phenotyped for key breeding fruit and plant traits at two locations over two seasons. A genome-wide association (GWA) analysis was performed using the mixed linear model, which takes into account both a kinship matrix and the sub-population membership of the accessions. Overall, 194 phenotype/genotype associations were uncovered, relating to 30 of the 33 measured traits. These associations involved 79 SNP loci mapping to 39 distinct chromosomal regions distributed over all 12 eggplant chromosomes. A comparison of the map positions of these SNPs with those of loci derived from conventional linkage mapping showed that GWA analysis both validated many of the known controlling loci and detected a large number of new marker/trait associations. Exploiting established syntenic relationships between eggplant chromosomes and those of tomato and pepper recognized orthologous regions in ten eggplant chromosomes harbouring genes influencing breeders' traits.

Linkage disequilibrium and genome-wide association analysis for anthocyanin pigmentation and fruit color in eggplant.

BACKGROUND: The genome-wide association (GWA) approach represents an alternative to biparental linkage mapping for determining the genetic basis of trait variation. Both approaches rely on recombination to re-arrange the genome, and seek to establish correlations between phenotype and genotype. The major advantages of GWA lie in being able to sample a much wider range of the phenotypic and genotypic variation present, in being able to exploit multiple rounds of historical recombination in many different lineages and to include multiple accessions of direct relevance to crop improvement. RESULTS: A 191 accessions eggplant (Solanum melongena L.) association panel, comprising a mixture of breeding lines, old varieties and landrace selections originating from Asia and the Mediterranean Basin, was SNP genotyped and scored for anthocyanin pigmentation and fruit color at two locations over two years. The panel formed two major clusters, reflecting geographical provenance and fruit type. The global level of linkage disequilibrium was 3.4 cM. A mixed linear model appeared to be the most appropriate for GWA. A set of 56 SNP locus/phenotype associations was identified and the genomic regions harboring these loci were distributed over nine of the 12 eggplant chromosomes. The associations were compared with the location of known QTL for the same traits. CONCLUSION: The GWA mapping approach was effective in validating a number of established QTL and, thanks to the wide diversity captured by the panel, was able to detect a series of novel marker/trait associations.

QTL mapping in eggplant reveals clusters of yield-related loci and orthology with the tomato genome.

In spite of its widespread cultivation and nutritional and economic importance, the eggplant (Solanum melongena L.) genome has not been extensively explored. A lack of knowledge of the patterns of inheritance of key agronomic traits has hindered the exploitation of marker technologies to accelerate its genetic improvement. An already established F2 intraspecific population of eggplant bred from the cross '305E40' x '67/3' was phenotyped for 20 agronomically relevant traits at two sites. Up to seven quantitative trait loci (QTL) per trait were identified and the percentage of the phenotypic variance (PV) explained per QTL ranged from 4 to 93%. Not all the QTL were detectable at both sites, but for each trait at least one major QTL (PV explained ≥ 10%) was identified. Although no detectable QTL x environment interaction was found, some QTL identified were location-specific. Many of the fruit-related QTL clustered within specific chromosomal regions, reflecting either linkage and/or pleiotropy. Evidence for putative tomato orthologous QTL/genes was obtained for several of the eggplant QTL. Information regarding the inheritance of key agronomic traits was obtained. Some of the QTL, along with their respective linked markers, may be useful in the context of marker-assisted breeding.

The population structure and diversity of eggplant from Asia and the Mediterranean Basin.

A collection of 238 eggplant breeding lines, heritage varieties and selections within local landraces provenanced from Asia and the Mediterranean Basin was phenotyped with respect to key plant and fruit traits, and genotyped using 24 microsatellite loci distributed uniformly throughout the genome. STRUCTURE analysis based on the genotypic data identified two major sub-groups, which to a large extent mirrored the provenance of the entries. With the goal to identify true-breeding types, 38 of the entries were discarded on the basis of microsatellite-based residual heterozygosity, along with a further nine which were not phenotypically uniform. The remaining 191 entries were scored for a set of 19 fruit and plant traits in a replicated experimental field trial. The phenotypic data were subjected to principal component and hierarchical principal component analyses, allowing three major morphological groups to be identified. All three morphological groups were represented in both the "Occidental" and the "Oriental" germplasm, so the correlation between the phenotypic and the genotypic data sets was quite weak. The relevance of these results for evolutionary studies and the further improvement of eggplant are discussed. The population structure of the core set of germplasm shows that it can be used as a basis for an association mapping approach.