Genomic selection for traits of interest in the EEAOC sugarcane breeding program

RACEDO, J; ROSSI, E; AYBAR GUCHEA, M; PEÑA MALAVERA, A.; BRUNO, C; PERERA, M. F.; NOGUERA, A; BONAMICO, N; BALZARINI, M; OSTENGO, S

Hyderabad

Congreso; ISSCT XXXI Congress; 2023

Sugarcane breeding is complex and slow, since the time taken from crossing to the release of commercial varieties takes 10-15 years of intensive work. This is due to the genetic complexity of the crop and the strong environmental influence on the characters of interest, mostly determined by quantitative effects. The progress of sugar yield per hectare has been mainly associated with cane yield; however, sucrose content in juice plays a fundamental economic role. A useful tool to increase the genetic gain rate consists of integrating phenotypic and genomic information in genomic selection (GS) procedures, and predicting the breeding value estimated from genomic data (genomic estimated breeding value, GEBV) for each genotype. The objective was to evaluate the feasibility of implementing GS in different stages of the local breeding program by characterising the prediction ability of GEBVs with different GS models and populations. Different GS models were examined on two populations of the Estación Experimental Agroindustrial Obispo Colombres breeding program: i) 88 advanced clones from a selection population (of the penultimate stage of the breeding scheme) genotyped with a Diversity Arrays Technology (DarT) chip, and ii) 182 accessions from the germplasm collection genotyped with DArTseq. Traits related to early maturity were evaluated during three harvesting ages in areas of influence for the program. GS models were fitted with the Bayesian Ridge Regression (RR), Bayes A, Bayes B and Bayes C methods. Efficiency was evaluated through the correlation between GEBVs estimated by each model and the BLUPs of the genotypic effect. Correlations were obtained by cross validation of the intra-population. Although there were small differences among the predictive capacities of different models, the GS efficiency depended mainly on the characteristic under study, being higher (r=0.43) for the GS model estimated by the RR method for the sugar-recovered trait in the germplasm collection group. Accuracies obtained for traits related to early maturation, a key characteristic for the region, are encouraging and support the selection of parents incorporating molecular information by estimating their GEBVs.