Kinetic parameter prediction using neural networks identifies limitations to C₄ photosynthesis

Area of Science:

• Plant Physiology
• Computational Biology
• Photosynthesis Research

Background:

• Kinetic models of photosynthesis are crucial for predicting plant traits and identifying limiting factors.
• Estimating hundreds of genotype-specific kinetic parameters for large-scale models is computationally challenging.
• Efficient methods are needed to bridge the gap between complex models and experimental data.

Purpose of the Study:

• To develop an efficient artificial neural network (C4TUNE) for predicting kinetic parameters of photosynthesis models.
• To enable rapid parameter estimation from photosynthesis response curves.
• To facilitate large-scale analysis of genotype-specific photosynthetic efficiency.

Main Methods:

• Developed C4TUNE, an artificial neural network trained on synthetic datasets of C4 photosynthesis model parameters and response curves.
• Utilized a surrogate neural network to accelerate C4TUNE training by predicting response curves from parameters.
• Validated C4TUNE's predictions by simulating the kinetic model with over 99% accuracy.

Main Results:

• C4TUNE accurately predicts genotype-specific kinetic parameters from minimal response curve data.
• The predicted parameters enable direct use in kinetic model simulations with excellent fit.
• Applied C4TUNE to 68 maize genotypes, identifying key factors limiting photosynthetic efficiency.

Conclusions:

• C4TUNE offers a fast and precise method for predicting photosynthesis model parameters.
• This approach significantly reduces the data requirements for large-scale kinetic modeling.
• C4TUNE facilitates the identification of genetic and environmental factors influencing photosynthetic performance.