Our research will improve the predictive ability of genomic selection (trait prediction from molecular marker genotypes) in forage grasses and legumes, oats and energy grasses with the aim of increasing the speed and efficiency of our breeding programmes for these crops.

 

Approach: We are building on and increasing the size of the training populations used for genomic selection to improve prediction accuracy. This will allow us to reduce the length of the breeding cycles and enable a more efficient identification of the best parents for hybrid production and polycross breeding scenarios. The breeding programmes for each of the crops will provide the source of appropriate and growing training populations, and the development of prediction models. Although the different reproductive biology of the crops mean that the strategies for genomic prediction vary between them, the objectives are common. Besides the use of genomic prediction in variety production, outcomes from other CSP themes will provide novel target traits, and identification of potential QTL via GWAS analyses.

 

Potential impact: Improving the predictive ability of genomic selection and demonstrating its utility in our crops will allow the development of legume, oat and forage and energy grass germplasm that can be used for variety generation including of ideotypes and provide research populations. Thus, our work will increase the precision of the breeding programmes, deliver a faster breeding cycle and identify optimal genotypes for future selection and hybridisation. The knowledge and tools resulting from the research will be used to develop future forage, Miscanthus and oat crops adapted to a range of climate change scenarios and end-uses.

 

Key research insights and findings: In Miscanthus we have used genomic prediction on selection indices to improve yield prediction1 and used genomic index selection to optimise multi-trait selection (e.g. of ideotypes)2. We are analysing the results of a statistically powerful 1000 genotype trial. The SNP markers for the genotypes have been called against the new miscanthus genome sequence resulting in 2.44 million SNPs across 19 chromosomes. SNPs markers have been used to identify 8 species groups and linkage disequilibrium studies (a control for GWAS and GP) have been done.

 

In oats, inter-generational and intra-generational genomic prediction models have been conducted using historical phenotype and genotype data. These are being expanded using both association mapping panels and current spring oat breeding programme populations that are currently being both phenotyped and genotyped. This will result in the development of multi-trait prediction models for key traits underpinning yield and grain quality and ultimately optimal parental identification for future hybridisation.

 

In oats work on genotyping-by-sequencing in oat genome research and its applications for both association analysis and genomic selection has been published (Bekele et al., 2018; dx.doi.org/10.1111/pbi.12888) and a further paper on the development of intra-generational prediction models is in preparation. This demonstrated that it is feasible to make within generation, between lineage genomic predictions in oats based on genotypic information. It also showed that early generation genotyping data can reduce the number of lines for later phenotyping based on selections of siblings to progress. Using sets of small families selected at an early generation could enable the use of genomic prediction for adaptation to multiple target environments at an early stage in the breeding program.

 

One paper has previously been published on genomic prediction in L. perenne (Grinberg et al., 2016; doi.org/10.3389/fpls.2016.00133). We have contributed to development of a potential framework for unifying genomic prediction in crops and livestock animals (dx.doi.org/10.1038/ng.3920), and a paper on genetic diversity and association analysis in red clover has been accepted in Scientific Reports pending revision.

 

In diploid perennial ryegrass, an outbreeding species, the breeding programme is based on population development through recurrent selection. We use phenotypic and genotypic data to predict the mean and variance of progeny derived from poly-crosses used in the breeding programme for selecting plants for the population improvement cycle, and in predicting the optimal parents in variety production. For example, the predicted performance of agronomic traits of a putative high lipid ryegrass variety was compared with other varieties using genomic breeding values of progeny from its four parents (Skøt et al. 2018; Grassland Science in Europe 23: 342-344). This tells us whether the novel variety was developed without predicted detrimental effects on agronomic performance. The work generates prediction models with improved accuracy including marker subsets which can be used to improve prediction of persistency and the performance of populations in response to biotic and abiotic stresses. We have increased the size of the training population to 830 genotypes, which has improved prediction ability to around 0.8 in the case of quality traits, and 0.5 for biomass yield. This is sufficient accuracy to allow us to use GS in the population improvement cycle. The genetic marker technology will change from

 

Four potential perennial ryegrass varieties have been made based on genomic breeding value which are undergoing phase 1 testing. More varieties of perennial ryegrass are in production on the basis of genomic breeding value, and we will propose a model for predicting progeny performance from parental crosses in the breeding programme. New oat varieties which combine favourable allele combinations are also in first year replicated field trials.

 

Note: If you are viewing this page on a mobile device the associated images will appear below.

 

> Publications

 

^ Top of current page

 

< Return to New Technologies