Polygenic prediction and gene regulation networks.

Area of Science:

• Systems Biology
• Quantitative Genetics
• Computational Biology

Background:

• Phenotypic variation arises from complex biological mechanisms.
• Statistical methods are crucial for understanding and predicting biological functions.
• Gene regulatory networks (GRNs) underpin complex traits.

Purpose of the Study:

• To assess the accuracy of statistical models in capturing phenotypic variation driven by nonlinear biological mechanisms.
• To integrate computational GRN models with linear prediction models.
• To explore the causal architecture of complex traits.

Main Methods:

• Developed a computational model of gene regulation networks.
• Created diverse populations of networks with identical topology but varying regulatory strengths.
• Employed a linear additive prediction model, similar to polygenic scores.
• Applied concepts from quantitative genetics and dynamical systems theory.

Main Results:

• Identified specific regulatory connections critical for phenotype prediction.
• Differentiated between core and peripheral causal determinants of complex traits.
• Established links to global sensitivity, local stability, and sloppy parameters.
• Investigated the role of epistasis in phenotype prediction.

Conclusions:

• Statistical models can effectively approximate phenotypic variation in complex biological systems.
• Understanding regulatory strengths is key to predicting complex traits.
• Findings inform the omnigenic model and highlight the importance of network structure.