Plant Genotype to Phenotype Prediction Using Machine Learning.

Area of Science:

• Agricultural Science
• Genetics
• Data Science

Background:

• Genomic prediction tools like genomic best linear unbiased prediction (GBLUP) are vital in crop breeding.
• Current methods struggle with non-linear relationships and high-dimensional data, such as UAV imagery.
• Machine learning (ML) offers potential for enhanced prediction accuracy.

Purpose of the Study:

• To review the application of statistical and ML methods for genotype to phenotype prediction in crop breeding.
• To discuss the potential and challenges of ML in this domain.
• To compare explainable models with ML approaches.

Main Methods:

• Literature review of statistical and machine learning algorithms.
• Analysis of methods for predicting phenotypic traits using genetic markers, environmental data, and imagery.
• Discussion of model explainability and data requirements.

Main Results:

• ML algorithms can autonomously extract features and model complex relationships, potentially improving prediction accuracy.
• Explainable models offer transparency but may lack predictive power for complex datasets.
• Challenges include data scarcity, inconsistent metadata, and ML model demands.

Conclusions:

• Machine learning holds significant potential to advance genotype to phenotype prediction in crop breeding.
• Addressing data quality and model interpretability are key to realizing ML's full potential.
• Further research is needed to integrate ML effectively into breeding programs.