Statistical Machine-Learning Methods for Genomic Prediction Using the SKM Library.

Area of Science:

• Plant breeding
• Computational biology
• Statistical genetics

Background:

• Genomic selection (GS) is a powerful predictive methodology transforming plant breeding.
• Successful GS implementation requires understanding statistical machine-learning (ML) methods.
• Breeders and scientists often lack the time and training for complex ML algorithms.

Purpose of the Study:

• To introduce state-of-the-art statistical ML methods for genomic prediction.
• To provide guidelines for implementing seven ML methods using the Sparse Kernel Methods (SKM) R library.
• To facilitate the use of advanced GS techniques by professionals without deep ML or programming expertise.

Main Methods:

• Utilized the Sparse Kernel Methods (SKM) R library.
• Provided implementation guidelines for seven ML methods: random forest, Bayesian models, support vector machine, gradient boosted machine, generalized linear models, partial least squares, and feed-forward artificial neural networks.
• Included functions for tuning, cross-validation, and performance evaluation metrics.

Main Results:

• Demonstrated the implementation of various ML methods for genomic prediction.
• Showcased the utility of the SKM R library for simplifying complex analyses.
• A toy dataset illustrated practical application for users with limited ML background.

Conclusions:

• The SKM R library enables breeders to implement advanced ML methods for genomic prediction without extensive programming knowledge.
• This approach democratizes the use of sophisticated predictive analytics in plant breeding.
• Simplified access to GS tools can accelerate breeding programs and improve crop development.