Genetic Analysis of Diallel Progeny Test Data Using Factor Analytic Linear Mixed Models
Multienvironmental trials are commonly used in plant breeding programs to select superior genotypes for specific sites or across multiple sites for breeding and deployment decisions. We compared the efficiency of factor analytic (FA) and other covariance structures for genetic analysis of height growth in Pinus taeda L. diallel progeny trials to account for heterogeneity in variances and covariances among different environments. Among the models fitted, FA models produced the smallest Akaike information criterion (AIC) model fit statistic. An unstructured (US) variance-covariance matrix produced a log likelihood value similar to that for the FA model but had a large number of parameters. As a result, some models with US covariance failed to converge. FA models captured both variance and covariance at the genetic level better than simpler models and provided more accurate predictions of breeding values. Narrow-sense heritability estimates for height from 10 different sites were about 0.20 when more complex variance structures were used, compared with 0.13 when simpler variance structures such as identity and block-diagonal variance structures were used. FA models are robust for modeling genotype × environment interaction, and they reduce the computational requirements of mixed-model analysis. On average, all 10 environments had additive genetic correlation of 0.83 and dominance genetic correlation of 0.91, suggesting that genotype × environment interaction should not be a concern for this specific population in the environments in which the genotypes were tested.
No Supplementary Data
No Article Media