Genomic prediction of heading date in intermediate wheatgrass (Thinopyrum intermedium)


Ragupathy R, Larsen JR, Frick M, Cattani DJ, Miller J, Elwood J, Zezula A, OrtegaPolo R, Soto-Cerda BJ, Laroche A. 2020. Genomic prediction of heading date in intermediate wheatgrass (Thinopyrum intermedium). PAG XXVIII Conference PE0510

Plain language summary

Latest results on using genemic data to identified markers related to flowering time in intermediate wheatgrass


In the changing climate scenario, perennializing annual crops is one of the prime strategies of sustainable crop production. Intermediate Wheatgrass (IW) is used as a source of ‘perenniality’ in crosses with annual wheat, and also being de novo domesticated as a new perennial cereal. Hence, there is a need to understand the genetic architecture of phenotypic traits contributing to perenniality. In this study, a diverse panel consisting of 276 individuals collected from The Land Institute and USDA-National Germplasm Resources Laboratory were evaluated for heading date at two Canadian locations, namely Lethbridge (2017, 2018 and 2019) and Winnipeg (2018). Phenotypic data was analysed using mixed model, and best linear unbiased predictor (BLUP) values of lines combined over years were estimated. Genotyping of accessions was carried by GBS and the sequencing data was processed by fast-GBS pipeline (Torkamaneh et al. 2017) using Thinopyrum intermedium draft assembly (Kantarski et al. 2017). Genotype calling and downstream quality filtering resulted in the identification of 73,401 genome-wide SNPs among 276 accessions. Population structure was studied using 18,763 random genome-wide SNPs in STRUCTURE v.2.3.4 (Pritchard et al. 2000) suggesting two populations with 243 and 33 accessions, respectively. The average coefficient of membership of individuals belonging to Population-1 was 0.872 with a percentage of shared alleles with Population-2 individuals of 0.128. The coefficient of population differentiation Fst was 0.12, which indicated a moderate population structure between populations. Five different genomic prediction models, namely ridge regression-best linear unbiased predictor (rrBLUP, Endelman 2011), Bayesian ridge regression (Bayesian rr; de los Campos et al. 2013), Bayes A and Bayes B (Meuwissen et al. 2001), and genomic best linear unbiased predictor (GBLUP; Habier et al. 2007) were evaluated with 80% of the lines as the training set, for estimating marker effects, and prediction accuracies ranged from 0.43 to 0.48. Genome wide association studies is being carried out to identify the genomic regions having significant association with heading date, and the results and biological insights will be presented.