Machine learning models for predicting morphological traits and optimizing genotype and planting date in roselle...

Area of Science:

• Agricultural Science
• Computational Biology
• Genetics

Background:

• Accurate prediction of Roselle morphological traits is crucial for improving crop productivity and environmental adaptability.
• Understanding genotype-environment interactions is key for effective crop breeding and management.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting key Roselle morphological traits.
• To identify optimal genotype and planting date combinations for maximizing trait values.

Main Methods:

• Utilized Random Forest (RF) and Multi-layer Perceptron (MLP) algorithms to model traits based on genotype and planting date.
• Generated dataset from field experiments with ten Roselle genotypes and five planting dates.
• Integrated RF model with Non-dominated Sorting Genetic Algorithm II (NSGA-II) for optimization.

Main Results:

• RF (R² = 0.84) outperformed MLP (R² = 0.80) in predicting morphological traits.
• Planting date was identified as a more significant factor influencing trait variation than genotype.
• Optimal conditions identified: Qaleganj genotype planted on May 5, yielding 26 branches, 176-day growth period, 116 bolls, and 1517 seeds per plant.

Conclusions:

• Machine learning, particularly RF, effectively models complex genotype-by-environment interactions in Roselle.
• Integrating ML with evolutionary algorithms like NSGA-II provides a powerful approach for crop trait optimization.
• Findings support the use of computational tools for enhanced crop improvement and informed agronomic decisions.