Decoding living systems: Reassessing crop model frontiers via biological dynamics and optimized phenotype

Area of Science:

• Agricultural Science
• Computational Biology
• Genetics

Background:

• Understanding genotype-by-environment interactions is crucial for optimizing crop phenotypic performance.
• AI models offer predictive accuracy but lack biological interpretability.
• Process-based models provide mechanistic insights but can be complex to optimize.

Purpose of the Study:

• To develop an inverse engineering framework for modeling and optimizing biological systems.
• To bridge the gap between computational optima and field-validated cultivars.
• To enable adaptation across regional precipitation gradients.

Main Methods:

• Implemented a three-layered framework: sensitivity analysis, genetic algorithm, and similarity analysis.
• Utilized 3 years of field characterization data for model validation.
• Employed Gaussian Mixture Model (GMM)-based classification for regional precipitation gradients.

Main Results:

• Sensitivity analysis identified key genetic coefficients for yield with robust rankings.
• Genetic algorithm revealed two adaptive strategies: extended growth for high water availability and shortened cycles for water deficit.
• Similarity analysis identified WAB56-50 and DKAP2 as top breeding candidates, with a 22-30% genetic gap to optima.

Conclusions:

• The framework successfully compressed crop evaluation and selection cycles.
• Identified specific cultivars and genetic improvements for enhanced yield and water-use efficiency.
• Demonstrated principles applicable across biological scales for modeling and trait measurement.