The main objective of this project is to develop genomic selection (GS) methods for breeding of the forage grass timothy. The project will utilize DNA-sequencing, bioinformatics and statistical methods to develop necessary tools for implementation of genomic selection in the timothy breeding programme at Graminor, the industrial partner in the project. GS is a breeding method that makes it possible to select parents for crossings based on their genotypic value, measured using genetic markers, instead of traditional selection based on phenotypic values.
In the last project reports we presented heritabilities, bias and accuracies calculated for all yield traits in the training population. The training population consist of 720 full-sib (FS) families tested in 3-years field trials, sown twice, at 2-3 locations in Norway (Løken, Staur and Vågønes 2003-2005; Løken, Vågønes and Arneberg (Graminor) 2006; and Løken and Arneberg 2007-2012). Overall, the GEBV predictions for dry matter yield has high correlations with the observed phenotypes, low bias and moderate heritabilities, as expected. GEBVs predicted in the training population using gBLUP models were compared with predictions developed by advanced machine learning models like Reproducing Kernel Hilbert Space (RKHS) in collaboration with Dr. Jose Crossa, CIMMYT, Mexico. Yield traits (dry matter yield (DMY)/cut/year, and total DMY (SDMY)/year and over years) have similar accuracies (0.950 to 0.997) detected by both gBLUP and machine learning models (RKHS). The GS models developed in the training population was implemented in the validation population for predicting GEBVs. The validation consists of 213 unrelated full-sib families (FS-families) with yield data recorded in 3-years field trials (2016, 2017 and 2018) at Bjørke (Graminor) and Løken. Correlations between the GEBVs for yield estimated in the validation population and the observed yield data were estimated in order to crosscheck the efficiency of the GS models. Some of the traits like KGDM201 (first cut, second year dry matter yield) and SUMDM2 (total dry matter yield, second year) have correlations of 0.45 and 0.29, respectively. However, for other traits, the correlations are very low. These results are similar as observed in other studies of genomic selection in forage grasses, e.g., perennial ryegrass.
In the present reporting period we have developed a new SNP marker set (180,000 SNPs) using a reference genome based on a rough draft genome of timothy developed using the Oxford Nanopore Sequencing Technology (in house). This new SNP markers are being used to predict GEBVs for yield and quality traits using the same prediction models developed and applied for the earlier SNP marker set. That marker set was based on a de novo genome assembly, i.e. short sequences and with low precision. We expect that the new SNP marker set will improve the correlations between the predicted breeding values and the observed yield data. The results of these analyses will be published in the coming year.
Dette er den første studien av genomisk seleksjon (GS) anvendt i timotei. Prosjektet viser lovende resultater for implementering av GS i foredlingsprogrammet hos Graminor, som har det nasjonale ansvaret for å foredle nye og forbedrede grassorter med høyt avlingspotensial, god kvalitet og tilpasning til norske vekstforhold. GS kan predikere genomisk-estimerte avlsverdier (GEBV) i timotei, spesielt for egenskaper der mange og presise fenotypiske observasjoner eksisterer. Genomsekvens for timotei, og unik kompetanse innen bioinformatikk og statistisk modellering for krysspollinerte arter er utviklet. Dette er avgjørende for vellykket implementering av GS i Graminor sine engvekstforedlingsprogrammer. GS muliggjør en raskere foredling av nye sorter, og vil øke konkurranseevnen til Graminor sine sorter på markedet. Landbruket vil ha stor nytte av forbedrede sorter med høy avkastning, god kvalitet og klimatilpasning som gir høyere inntekt i melk- og kjøttproduksjonen.
The primary objective of this project is to develop genomic tools for breeding of high quality timothy cultivars adapted to Norwegian growing conditions. Timothy is one of the most important input factors for Norwegian farming, and new improved cultivars with proper adaptation to the future climate will improve the economic value of milk and meat production. In this project, we will employ 'state-of-the art' genotyping and bioinformatic methods to develop tools that can be used to implement genomic selection (GS) in the breeding programme of timothy at Graminor AS, the industrial partner in this project. A large number of full-sib families which has been phenotyped for yield and forage quality in progeny tests will be genotyped using genotyping-by-sequencing (GBS) methods. Family-based genome-wide allele frequency profiles will be derived and used in conjunction with the phenotypic data to study associations and develop different GS models. The predictive power of different GS models will be tested, and finally validated in an unrelated set of families which are being progeny tested during the course of the project. Strategies for implementation of GS breeding for timothy will be developed in collaboration with the Danish partners. Graminor AS has a relatively large breeding activity in timothy with the aim of capturing market shares also in markets outside Norway. Development and implementation genomic-based breeding methods will be a key element for the competitive strength of Graminor AS in the near future. For NMBU this project is of great academic interest.