Machine learning approaches for crop improvement: Leveraging phenotypic and genotypic big data.

Area of Science:

• Agricultural Science
• Genetics
• Computational Biology

Background:

• Traditional crop breeding relies heavily on phenotyping, which is resource-intensive and time-consuming.
• Genomic selection (GS) offers a more efficient alternative by utilizing genomic data to predict breeding values.
• Advances in machine learning (ML) and genotyping technologies have accelerated GS applications.

Purpose of the Study:

• To systematically review ML approaches for genomic selection in major crops over the last decade.
• To identify factors influencing GS model performance and transferability across environments.
• To explore future directions for integrating omics data and biological networks into crop improvement strategies.

Main Methods:

• Systematic literature review of ML applications in genomic selection for crop traits.
• Analysis of studies focusing on single and multiple trait selection.
• Critical assessment of factors affecting model accuracy and environmental transferability.

Main Results:

• ML approaches have shown significant potential in improving the accuracy and efficiency of genomic selection for various crop traits.
• The review identified key ML algorithms and their performance in different crop species and breeding scenarios.
• Data on intermediate phenotypes (metabolites, gene expression) and advanced modeling techniques are crucial for further GS improvements.

Conclusions:

• Genomic selection, powered by machine learning, is revolutionizing crop breeding by shortening cycles and reducing phenotyping costs.
• Integrating high-throughput omics data and biological networks with ML models holds immense promise for future crop improvement.
• Further research is needed to enhance model transferability across diverse environments and optimize data integration strategies.