bioRxiv2023
Plant growth and resilience is greater than the sum of its component traits, with important traits influencing one another. This interdependency makes it challenging to identify the genetic determinants of key agronomic traits, such as water use efficiency. Measuring traits such as plant height and size is possible via image analysis software, such as PlantCV, but these traits are often highly correlated. Furthermore, plant size is estimated using a 2D projection of a 3D object, which is more difficult for plants with complex body plans. To address these problems, we developed a generalizable, biologically-informed model describing the temporal coordination of semi-sequential phytomeric growth in the model grass Setaria. Our approach integrates time dependence with water usage, the growth of phytomers, and the emergence of side shoots or tillers, improving our ability to estimate both phytomer-level phenotypes and water usage traits in high-throughput. We developed new PlantCV methods to inform the Phytomeric Growth Model, and estimated parameter traits using a high-throughput phenotyping dataset of a recombinant inbred line (RIL) panel in well-watered and drought conditions. Model parameter estimates identified additional quantitative trait loci (QTL) for new traits compared to the directly measured PlantCV-derived estimates, and predicted the relationships underlying QTL control of composite traits through co-localization of model parameters to loci controlling size and water use metrics. The Phytomeric Growth Model estimates improved our ability to estimate the growth of tillers and obtain a model-based estimate of marginal water use efficiency (defined here as the ratio of biomass gained per gram of water transpired), both identified as highly influential on plant drought responses through random forest classification. Our approach demonstrates the value of integrating mechanistic modeling with high-throughput imaging to extract new information at scale.