Reconstruction of Networks with Direct and Indirect Genetic Effects.

Area of Science:

• Quantitative genetics
• Causal inference
• Plant breeding

Background:

• Understanding genetic variance requires distinguishing direct and indirect genetic effects on phenotypic traits.
• Current causal inference methods often struggle with 'missing heritability' and cannot infer direct genetic effects.
• Existing multi-trait mixed models (MTM) are computationally challenging and limit direct genetic effect inference.

Purpose of the Study:

• To propose an alternative strategy for causal inference that formally includes genetic effects in the graph.
• To develop a method that can analyze a larger number of traits and test for direct genetic effects.
• To improve the accuracy of reconstructing trait relationships using individual plant or plot data.

Main Methods:

• Developed the PCgen algorithm, which integrates genetic effects directly into causal inference graphs.
• PCgen allows for the analysis of multiple traits simultaneously and enables testing for direct genetic effects.
• Utilized individual plant or plot data for more accurate reconstruction of trait relationships compared to genotypic means.

Main Results:

• The PCgen algorithm successfully incorporates genetic effects into causal inference, overcoming limitations of previous methods.
• The method allows for direct testing of direct genetic effects and improves the orientation of edges between traits.
• Reconstruction of trait relationships is significantly more accurate when using individual-level data.

Conclusions:

• PCgen provides a powerful new approach for dissecting complex genetic architectures and understanding trait relationships.
• The ability to test for direct genetic effects and handle multiple traits opens new avenues for genetic research.
• The PCgen algorithm, implemented in the R-package pcgen, offers a valuable tool for quantitative geneticists and breeders.