Inferring a directed acyclic graph of phenotypes from GWAS summary statistics

Area of Science:

• Computational biology
• Genetics
• Network analysis

Background:

• Understanding disease etiology is crucial for developing effective treatments.
• Phenotype network estimation is an emerging computational biology field.
• Existing methods often require individual-level genetic data.

Approach:

• Developed a method to construct phenotype networks using Gaussian linear structure models and directed acyclic graphs (DAGs).
• Utilized genetic variants as instrumental variables and summary statistics from genome-wide association studies (GWAS).
• Incorporated a summary statistics-based likelihood ratio test for directed edges.

Key Points:

• The method requires only summary statistics and reference genotype data, enhancing accessibility.
• Successfully estimated a causal network of 29 cardiovascular-related proteins.
• Linked the cardiovascular network to Alzheimer's disease (AD) etiology.
• Demonstrated method effectiveness through simulation studies.

Conclusions:

• The proposed method provides an efficient way to estimate phenotype networks from GWAS summary statistics.
• Facilitates deeper understanding of complex disease mechanisms and genetic architectures.
• The R package 'sumdag' and associated resources are publicly available for broader application.