Maximizing genetic gain for rice (Oryza sativa L.) grain yield by implementing genomic selection

Improving a quantitative trait like grain yield in rice using conventional breeding approaches is time and resource-demanding. Utilizing genomic selection for improving grain yield in rice is assumed to be promising. A founder population genotyped with novel genomic markers was used as a training population. The training population was phenotyped over three years for grain yield. A bi-parental population developed from parents selected from the training population genotyped with the same markers was used as the testing population. Four different predictive models were used on the training population at different marker densities. The results indicated that lower marker densities leads to poor predictive abilities among all models. Increasing marker density improves the prediction ability; however, the increment in predictive ability over the mid-density of markers was relatively low. The candidate genotypes selected based on predicted performance in the testing population showed a 20% higher genetic gain over the testing population mean, a 16% higher gain over the training population mean, and a 150% higher gain over the mid-parent value. The mid-density markers uniformly covering the rice genome uniformly are sufficient to implement genomic selection in rice. Integrating genomic selection into ongoing breeding programs would benefit the breeder in selecting potential candidates for improving grain yield in rice.