Prediction accuracy of genomic selection models for earliness in tomato

Genomic selection is considered to be an important tool in plant breeding programs. However, its application in the earliness of tomato (Solanum lycopersicum L.) has not been studied. The objective of the present study was to evaluate the prediction performance of six statistical models for six quantitative characteristics related to earliness in tomato. The study used phenotypic and genotypic data belonging to an [F.sub.2] population consisting of 172 tomato plants. Simple sequence repeat (SSR) markers were obtained using genotypic information, and the genomic values were estimated by the following six different statistical models: Bayesian Lasso (BL), Bayesian ridge regression (BRR), BayesA, BayesB, BayesC[pi], and reproducing kernel Hilbert spaces (RKHS) regression. The correlation values ranged from 0.17 to 0.57. The highest association values were found in days to flowering of the third inflorescence and 1000-seed weight, which were greater than 0.5. In general, all the models performed in a similar manner because only slight differences were observed among the correlation values. Specifically, BL, BayesB, and RKHS exhibited the highest Pearson correlation values for most traits. According to the results, genomic selection could be a useful tool to support tomato breeding focused on earliness. Key words: Genetic gain, genomic selection, Solanum lycopersicum, statistical models.