Can Genomic Selection Help Chickpea Breeding to Develop Superior Lines with Higher Yield Under Drought Stress?

Genomic selection (GS) unlike MABC (marker assisted backcross) predicts breeding values of lines using genome-wide marker profiling and makes selection prior to phenotyping. To address the low productivity in chickpea owning to several biotic and abiotic stresses, a collection of 320 elite breeding lines was selected as the “training population”. Training population was phenotyped extensively for yield and yield related traits at ICRISAT, Patancheru and IARI, Delhi during crop season 2011-12, and 2012-13 under rain-fed and irrigated conditions. In parallel, the training population was genotyped using DArT arrays (15,360 features). Phenotypic data along with genome-wide marker profiling data (974 polymorphic markers) were used with six statistical models including RR-BLUP, kinship based ridge regression, BayesCπ, BayesB, Bayesian LASSO and random forest to predict genomic estimated breeding values (GEBVs). GS models were tested for four yield related traits namely seed yield, 100 seed weight, days to 50% flowering and days to maturity. Correlation inside training (CIT) for the models tested varied from 0.138 to 0.912. Heat map analysis using genotyping data to understand the relationship within these lines suggested possibility of two different groups similar to cluster analysis. In order to understand the effect of population structure on accuracy, analysis was re-performed by implementing population structure for calculation of GEBV. Population structure significantly affected the CIT that varied from 0.001 to 0.745 for desi group, and 0.004 to 0.727 for kabuli group. Lines with higher GEBVs for each trait have been selected and are being used for crossing.