A Plant Breeding, Genetics, and Biotechnology Division Seminar
By Fang Wang
Visiting research fellow
Crop Research Informatics Laboratory, IRRI
2:00 P.M., Wednesday, 15 May 2013
Room A, D.L. Umali Laboratory, IRRI
Genomic selection (GS) is a new marker-assisted breeding method. It uses breeding (genotypic) values estimated using molecular markers spanning entire genome as selection criterion. GS has the potential to increase genetic gain by reducing breeding cycle and/or increasing selection accuracy and can greatly reduce the cost of phenotyping. Accurate prediction of genomic breeding (genotypic) values (GEBVs) presents a central challenge to contemporary plant breeders. A wide array of statistical methods has been developed for estimating GEBVs. The relative prediction abilities of various methods are trait and population specific. For any datasets many different methods must be tested to develop the best predictive model. As a first step towards the deployment of GS in rice breeding, several GS methods were applied to a population of 392 rice advanced lines tested in 8 environments for 4 traits and genotyped using 105 SSR markers. The statistical methods tested methods included the pedigree-based genetic relatedness BLUP (P-BLUP), marker-based genetic relatedness BLUP (G-BLUP), random forests (RF), reproducing kernel Hilbert spaces (RKHS), Bayesian LASSO (BL) and Bayesian Ridge Regression (BRR). It was found that the best GS models could have prediction ability higher than 0.5 for all the 4 traits in all 8 environments, which was very promising. The challenges of implementing GS and our planned future activities will also be discussed.