GS + de novo GWAS in Tropical and Temperate Irrigated Rice Breeding Programs

GS + de novo GWAS in tropical and temperate irrigated rice breeding programs

Spindel, J.E.^a, Monteverde, E.^a, Begum, H.^b,d., Akdemir, D.^a, Collard, B.^b, Redoña, E.^b,e, Blanco, P. ^f, Pérez de Vida, F. ^f, Bonnecarrère, V. ^g, Gutiérrez, L. ^h, Rosas, J. ^f,  Quero, G. ^h, Berberián, N. ^h, Garaycochea, S. ^g, Fernández, S. ^g, Jannink, J.L.^a,c, McCouch, S.^a

^a Department of Plant Breeding and Genetics, 240 Emerson Hall, Cornell University, , Ithaca, NY, 14853. 

^b International Rice Research Institute, Los Baños, Philippines. (Mailing address DAPO Box 7777, Metro Manila 1301, Philippines).

^c USDA-ARS, North Atlantic Ares, Robert W. Holley Center for Agriculture and Health, 538 Tower Road, Ithaca, NY 14853

^d Current Address: Bangladesh Rice Research Institute, Gazipur-1701, Bangladesh

^e Current Address: Delta Research and Extension Center, 82 Stoneville Rd., P.O. Box 197, Stoneville, MS 38776

^f Instituto Nacional de Investigación Agropecuaria-Treinta y Tres. Route 8, Km 282, 33000 Treinta y Tres, Uruguay.

^g Instituto Nacional de Investigación Agropecuaria-Las Brujas. Route 48, Km 10, Rincón del Colorado, 90200, Canelones, Uruguay.

^h Facultad de Agronomía, Universidad de la República, Garzón 780,  12900, Montevideo, Uruguay.

Genomic selection (GS) helps accelerate the rate of genetic gain in breeding by utilizing whole genome data to predict the breeding value of offspring. Genomic selection has been widely applied to maize and small grains breeding, but has only recently been applied to rice. Here, we discuss the results of GS cross-validation in two rice breeding programs: 1. the International Rice Research Institute (IRRI) irrigated rice breeding program for Southeast Asia (tropical rice), and 2. the National Institute for Agronomic Research in Uruguay (INIA) irrigated rice breeding program (temperate rice).  Both breeding populations were genotyped using genotyping-by-sequencing (GBS) and genomic prediction performed for grain yield, plant height, and flowering time. In addition, prediction was performed in the INIA population for milling yield and grain chalkiness. RR-BLUP accuracies in both populations averaged ~0.3 for grain yield, however, by combining RR-BLUP with markers fit as fixed effects selected from the results of a genome-wide-association study (GWAS) on the RR-BLUP training data in the IRRI population we improved prediction accuracies up to 30% depending on the trait. We also show how making careful use of multi-environment data can improve prediction accuracies up to 10-fold over single environment models.