A0840
Title: Modelling soybean growth: A nonlinear mixed model approach
Authors: Maud Delattre - INRAE (France) [presenting]
Hiroyoshi Iwata - The University of Tokyo (Japan)
Jessica Tressou - INRAE (France)
Abstract: Field experiments on soybean were conducted in Arid Land Research Center, Tottori, Japan, under several experimental conditions. The growth was monitored by a drone measuring each day the plant height of about 200 soybean varieties for which whole-genome sequence data are also available. Based on these data, the objective is to propose an original statistical approach to refine the understanding of the determinants of soybean growth and improve the prediction of phenotypic traits of interest. The problem is formalized through a nonlinear mixed-effects model in which random effects allow modeling of genetic and environmental effects and their variability. Parameter estimation in nonlinear mixed models is however not straightforward, especially due to the model's nonlinearity and the random effects. SAEM (Stochastic Approximation of the Expectation-Maximization algorithm) is widely used in this context, but it is rarely used in plant biology. The originality compared to standard mixed-effects models is that the soybean model integrates the relationships between varieties through the kinship matrix, which requires an adaptation of the algorithm. SAEM is implemented and predictions of expected growth curves per variety can then be deduced by a maximum a posteriori approach. The methodology is applied to the experimental data from Tottori.