Phenotypic trait identification using a multimodel Bayesian method: A case study using photosynthesis in Brassica rapa genotypes

Abstract

Agronomists have used statistical crop models to predict yield on a genotype-by-genotype basis. Mechanistic models, based on fundamental physiological processes common across plant taxa, will ultimately enable yield prediction applicable to diverse genotypes and crops. Here, genotypic information is combined with multiple mechanistically based models to characterize photosynthetic trait differentiation among genotypes of Brassica rapa. Infrared leaf gas exchange and chlorophyll fluorescence observations are analyzed using Bayesian methods. Three advantages of Bayesian approaches are employed: a hierarchical model structure, the testing of parameter estimates with posterior predictive checks and a multimodel complexity analysis. In all, eight models of photosynthesis are compared for fit to data and penalized for complexity using deviance information criteria (DIC) at the genotype scale. The multimodel evaluation improves the credibility of trait estimates using posterior distributions. Traits with important implications for yield in crops, including maximum rate of carboxylation (Vcmax) and maximum rate of electron transport (Jmax) show genotypic differentiation. B. rapa shows phenotypic diversity in causal traits with the potential for genetic enhancement of photosynthesis. This multimodel screening represents a statistically rigorous method for characterizing genotypic differences in traits with clear biophysical consequences to growth and productivity within large crop breeding populations with application across plant processes.