Seleção Genômica Ampla (GWS) sob assimetria para resistência à podridão da espiga em milho

Abstract

Maize is a crop of great economic impact, but has its productivity affected by the fusarium verticilioides pathogen, which can cause rotten kernels and mycotoxins. In addition to all the management that must be done to control this disease, the use of resistant genotypes is the most effective. Several studies report that resistance to these diseases is controlled by genes of quantitative inheritance, and phenotypic selection is difficult in these characters, due to low heritability and high influence of the environment. Among the most used tools in plant breeding programs, the Wide Genomic Selection (GWS) is highly effective in selecting superior genotypes. Some characters of quantitative character may present skew normal distribution, mainly on resistance to plant diseases. When this occurs, data transformation is not always an effective alternative, and the use of models that deal with this skew normal is recommended. Therefore, this work aimed to verify the efficiency in the use of Mixed Normal Asymmetric Bayesian Model in the prediction of data with skew normal distribution and by GBLUP. Phenotypic analyzes were performed in the Lavras and Uberlândia environments and three characters were evaluated: percentage of rotten kernels, proportion of diseased ears and ear rot score. After verifying the data, the transformation was made as a way to correct non-normality, but even so the data presented skew normal distribution. In the analysis of the estimated parameters, the characters rotten kernels and score showed greater heritability compared to the proportion of diseased ears, so these characters can be used to obtain genotypes resistant to ear rot caused by fusarium verticilioides. In the analyzes with the GBLUB and the Bayesian Asymmetric Model, a high heritability and correlation were observed for the characters analyzed under the Bayesian Asymmetric Model, different from GBLUP, which obtained a lower heritability and less correlation. The high correlation and good genomic prediction presented by the Bayesian Asymmetric Model leads to infer that this model is effective in analyzing data with asymmetric distribution.