Heritability estimation for a linear combination of phenotypes via ridge regression

Area of Science:

• Quantitative genetics
• Genomic prediction
• Statistical genomics

Background:

• Joint analysis of multiple phenotypes is crucial in breeding programs.
• Current heritability estimation methods often focus on single traits and have limitations.
• There is a need for more flexible and interpretable heritability estimation methods.

Purpose of the Study:

• To propose a novel heritability estimator for linear combinations of multiple phenotypes.
• To develop a method that accounts for correlations among phenotypes.
• To provide a flexible and interpretable alternative to existing heritability estimation techniques.

Main Methods:

• Utilized multivariate ridge regression for heritability estimation.
• Developed an estimator for linear combinations of phenotypes.
• Investigated performance in sparse and dense high-dimensional settings.

Main Results:

• The proposed estimator provides accurate heritability estimates in both sparse and dense scenarios.
• The method is consistent and asymptotically normally distributed under mild conditions.
• Demonstrated improved performance over independent heritability estimates by incorporating trait correlations.
• Applied the method to the Oryza sativa rice dataset for heritability and correlation analysis.

Conclusions:

• The novel multivariate ridge regression estimator offers a significant improvement for heritability estimation with multiple phenotypes.
• The method effectively leverages correlation information among traits.
• An R package is available for practical implementation.