Mapping and Predicting Non-Linear Brassica rapa Growth Phenotypes Based on Bayesian and Frequentist Complex Trait Estimation

Predicting phenotypes based on genotypes and understanding the effects of complex multi-locus traits on plant performance requires a description of the underlying developmental processes, growth trajectories, and their genomic architecture. Using data from Brassica rapa genotypes grown in multiple density settings and seasons, we applied a hierarchical Bayesian Function-Valued Trait (FVT) approach to fit logistic growth curves to leaf phenotypic data (length and width) and characterize leaf development. We found evidence of genetic variation in phenotypic plasticity of rate and duration of leaf growth to growing season. In contrast, the magnitude of the plastic response for maximum leaf size was relatively small, suggesting that growth dynamics vs. final leaf sizes have distinct patterns of environmental sensitivity. Consistent with patterns of phenotypic plasticity, several QTL-by-year interactions were significant for parameters describing leaf growth rates and durations but not leaf size. In comparison to frequentist approaches for estimating leaf FVT, Bayesian trait estimation resulted in more mapped QTL that tended to have greater average LOD scores and to explain a greater proportion of trait variance. We then constructed QTL-based predictive models for leaf growth rate and final size using data from one treatment (uncrowded plants in one growing season). Models successfully predicted non-linear developmental phenotypes for genotypes not used in model construction and, due to a lack of QTL-by-treatment interactions, predicted phenotypes across sites differing in plant density.

https://ift.tt/2GB6KBL