Deep-learning power and perspectives for genomic selection

Area of Science:

• Artificial Intelligence
• Genomics
• Plant Breeding

Background:

• Deep learning (DL) has surpassed human capabilities in tasks like image classification.
• DL models offer powerful solutions for complex prediction tasks.
• Genomic selection (GS) is a key methodology in modern plant breeding.

Purpose of the Study:

• To explore the potential of DL in accelerating GS.
• To highlight how DL can enhance prediction accuracy in plant breeding.
• To identify opportunities and challenges in applying DL to GS.

Main Methods:

• Review of DL applications in artificial intelligence and machine learning.
• Analysis of DL model performance in complex prediction tasks.
• Discussion on the integration of DL with GS methodologies.

Main Results:

• DL models demonstrate superior performance compared to traditional methods in various AI tasks.
• The integration of DL with GS shows promise for advancing plant breeding.
• Current DL methods require adaptation to effectively utilize unique breeding data.

Conclusions:

• DL holds significant power to revolutionize plant breeding through enhanced GS.
• Further research is essential to develop specialized DL methods for breeding applications.
• Optimizing DL for breeding inputs is crucial for realizing its full potential in GS.