A DeFries and Fulker regression model for genetic nonadditivity

Area of Science:

• Behavioral Genetics
• Quantitative Genetics
• Twin Studies

Background:

• The DeFries and Fulker (DF) regression method is widely used in behavioral genetics for twin studies.
• The standard DF model can yield biologically implausible parameter estimates (e.g., heritability > 1 or < 0).
• These issues arise when the correlation between monozygotic twins exceeds twice the correlation between dizygotic twins, or with genetic nonadditivity.

Purpose of the Study:

• To demonstrate the biases in heritability and shared environmentality estimates from the original DF model when genetic nonadditivity is present.
• To introduce a novel regression equation for unbiased estimation of additive and dominance genetic variance components.
• To evaluate the performance of a constrained DF model and compare its estimates with maximum-likelihood procedures.

Main Methods:

• Algebraic derivation of bias in the original DF model due to genetic nonadditivity.
• Development of a simple regression equation for standardized additive and dominance genetic variance components.
• Application of the methods to a large dataset (6 million twin pairs) from the Monte Carlo Twin Registry.
• Derivation of expectations for a constrained DF model.

Main Results:

• The original DF model produces positively biased heritability and negatively biased shared environmentality estimates under genetic nonadditivity.
• The proposed regression equation yields unbiased estimates for additive and dominance genetic variance.
• Parameter estimates from the revised DF method closely match those from maximum-likelihood procedures.
• The constrained DF model yields negatively biased estimates of broad-sense heritability.

Conclusions:

• The original DF regression model requires careful application, particularly when genetic nonadditivity is suspected.
• The proposed regression equation provides a more accurate method for estimating genetic variance components in twin studies.
• Maximum-likelihood procedures and the revised DF regression approach offer reliable estimates of genetic parameters.
• The constrained DF model is not suitable for estimating broad-sense heritability accurately.