Inferring genetic values and variances from pre-processed phenotypes

Area of Science:

• Quantitative genetics
• Statistical genetics
• Animal breeding

Background:

• Phenotypic data often requires pre-processing to simplify analysis.
• Observational field data, especially for complex traits, is frequently adjusted for nuisance factors like location and environment.
• These adjustments aim to reveal underlying genetic values masked by environmental influences.

Purpose of the Study:

• To investigate the consequences of omitting pre-correction for nuisance (fixed) effects in genetic analyses.
• To evaluate how pre-correction impacts the prediction of random effects with varying data density.
• To assess the effects of pre-correction on variance component estimation methods.

Main Methods:

• Utilized a mixed linear model framework for analysis.
• Employed real and synthetic datasets from plant and animal breeding.
• Examined variance component estimation using ANOVA, MINQUE, Maximum Likelihood, and Bayesian methods.
• Illustrated effects using a Braford cattle birth weight dataset.

Main Results:

• Ignoring pre-correction can change model rank and potentially distort genetic inferences.
• The impact of omitting pre-correction is contingent upon the interplay between model complexity and sample size.
• Pre-corrected data can affect the prediction of random effects, particularly with sparse fixed-effect information.

Conclusions:

• Pre-processing of phenotypic data must be applied cautiously in genetic inference.
• Failure to account for nuisance factors can alter data structure and complicate biological interpretation.
• The decision to pre-correct should consider the balance between computational simplification and potential inferential distortion.