Inference about quantitative traits under selection: a Bayesian revisitation for the post-genomic era

Area of Science:

• Quantitative genetics
• Statistical genetics
• Bioinformatics

Background:

• Selection schemes can distort genetic inferences and predictions, particularly in breeding and trait association studies.
• Biased inferences arise when data are not collected from random samples.
• Existing methods may not adequately address distortions caused by selection processes.

Purpose of the Study:

• To revisit inference in quantitative genetics under selection processes.
• To develop a unified Bayesian framework for inference and prediction that accounts for selection.
• To explore the impact of selection on genetic analyses in animal and plant breeding.

Main Methods:

• Integration of classical fitness concepts with missing data techniques.
• A fully Bayesian approach for unified inference and prediction.
• Development of a flexible "soft selection" model to diagnose selection's impact.

Main Results:

• The "soft selection" model helps assess the extent to which selection can be ignored.
• Highlights the link between missingness probability and fitness in selection scenarios.
• Demonstrates that a fixed selection threshold is often unrealistic; soft selection accounts for data variability.

Conclusions:

• The proposed Bayesian approach offers an integrated solution for inference and prediction under selection.
• While the quality of inferences under selection remains challenging to ascertain unambiguously, predictions can be empirically validated.
• The methods are applicable to natural selection and complex breeding data structures.