BLUP assessment of maize single-crosses performance using similarity in state and relationship coefficients

Whereas pedigree information is not always known in plant breeding, it is increasingly common the use of molecular markers to estimate the relationship used genomic selection studies, leading to biased estimates of kinship. This study aimed to investigate the effect of similarity in state and relationship coefficients in the assess- ment of single-crosses performance, and to verify the efficiency of genomic selection using these coefficients in identifying superior maize single-crosses. Five inbred lines were crossed to other 50 inbred lines to obtain 250 maize single-crosses that were evaluated for grain yield in 13 environments. 614 AFLP markers were used to ob- tain the molecular matrices used in the estimation of Queller and Goodnight, Ritland, Lynch and Rytland, Wang, and VanRaden relationship coefficients, and Jaccard, Sorensen-Dice, and Modified Rogers similarity coefficients, subsequently used in a mixed model based on additive and dominance relationship matrices for the assessment of maize single-crosses performance. Then, it was performed the phenotypic selection (PS) and the genomic se- lection (GS) of the 25 best single-crosses using each coefficient. Results showed that the conceptual difference between similarity and relatedness coefficients do not have much importance in the context of plant breeding, since both approaches do not consider single-crosses genealogy, but only their molecular information. The mod- els using Wang and Modified Rogers coefficients were superior both in assessing single-crosses performance and in performing genomic selection, suggesting the potential of these approaches to be used in maize breeding programs.