CORESTA Meeting, Agronomy/Phytopathology, 2023, Cancun, APPOST 04

Estimation and prediction of genetic parameters and breeding values through REML/BLUP approach

(1) Alliance One International, Global Center of Research, Development & Deployment, Brazil; (2) University of Arkansas, Crop, Soil and Environmental Sciences Department, U.S.A.

The Mixed Linear Models provides advantages compared to ordinary linear models. One is the ability to consider variables as random and other variables as fixed. The objective of this analysis is for the selection of superior genotypes. As the genotypes are a random sample of the hybrids within the trial, the genotypic effects are assumed as random and the consequent use of best linear unbiased predictor (BLUP) is justified. The objective was to predict the genetic value of tobacco hybrids through their BLUP, using Mixed Linear Models and parameters estimation by the Restricted Maximum Likelihood method (REML), for purposes of selection. The performance of 14 flue-cured Virginia tobacco hybrids, developed by Alliance One International, was studied. Yield, nicotine, sugar and quality index grade were evaluated across 11 environments in the southern region of Brazil. The trials were conducted in a randomized complete blocks design, with three replications and four commercial controls. Statistical analyses were performed in the JMP software program (SAS Institute Inc.), using Mixed Linear Models. Genetic parameters were estimated via REML, with genotypic means adjusted and estimated using the BLUP procedure. The likelihood ratio test (LRT) was performed for the variables evaluated in the experiment, and the significance was verified by the Chi-Square test. The use of the statistical approach of Mixed Linear Models showed to be effective in the selection of superior genotypes. The G × E (genotype x environment) interaction had a statistically significant impact on the genotypic and phenotypic parameters associated with tobacco yield, nicotine and quality index. It was possible to identify genotypes adapted to certain regions of the study, as well as widely adapted. This statistical analysis methodology allowed a better estimate of the genotypic values, making the selection and decision-making process more efficient on the part of the breeders.