Reproducing kernel Hilbert spaces regression: a general framework for genetic evaluation.

Area of Science:

• Statistical Learning
• Quantitative Genetics
• Bioinformatics

Background:

• Reproducing kernel Hilbert spaces (RKHS) methods are established statistical learning tools.
• RKHS methods are increasingly explored for integrating dense genetic markers into models.
• Existing genetic evaluation models often lack a unified theoretical framework.

Purpose of the Study:

• To present RKHS regression as a general framework for genetic evaluation.
• To demonstrate the adaptability of RKHS to diverse genetic models and data types.
• To show that standard genetic models are special cases within the RKHS framework.

Main Methods:

• Utilizing RKHS regression for statistical genetic analysis.
• Applying the framework to both pedigree-based and marker-based regression scenarios.
• Formulating genetic models within the RKHS context, including infinitesimal and non-infinitesimal, additive and non-additive effects.

Main Results:

• RKHS regression provides a unified approach applicable to various genetic evaluation scenarios.
• The framework accommodates both pedigree and dense marker data seamlessly.
• Standard genetic models, such as infinitesimal animal/sire models and marker-assisted selection, are shown to be specific instances of RKHS regression.

Conclusions:

• RKHS regression offers a powerful and flexible general framework for genetic evaluation.
• This unified approach simplifies and potentially enhances the analysis of complex genetic data.
• The RKHS framework has broad implications for statistical learning in genetics and beyond.