Estimation of complex effect-size distributions using summary-level statistics from genome-wide association studies...

Area of Science:

• Genetics and Genomics
• Statistical Genetics
• Human Complex Traits

Background:

• Genome-Wide Association Studies (GWAS) have identified numerous genetic variants associated with complex traits.
• Understanding the distribution of effect sizes for these Single Nucleotide Polymorphisms (SNPs) is crucial for interpreting GWAS findings.
• Polygenicity, the involvement of many genes with small effects, is a common feature of complex traits.

Purpose of the Study:

• To develop a statistical framework for estimating effect-size distributions of common variants using summary statistics and linkage disequilibrium (LD) information.
• To characterize the degree and nature of polygenicity across diverse human traits.
• To project the sample size requirements for future genetic studies and the predictive ability of polygenic risk scores (PRS).

Main Methods:

• Developed a likelihood-based method to analyze summary-level GWAS data and external LD information.
• Modeled effect-size distributions using a mixture of normal distributions, incorporating the proportion of causal SNPs.
• Applied the method to results from 32 diverse GWAS, including psychiatric, chronic disease, and anthropometric traits.

Main Results:

• All analyzed traits are highly polygenic, but exhibit substantial variation in genetic architecture.
• Psychiatric diseases and traits related to mental health/ability show the highest degree of polygenicity, characterized by a continuum of small effects.
• Other traits, such as chronic diseases, display distinct clusters of SNPs with varying effect magnitudes.

Conclusions:

• The findings highlight the diverse genetic underpinnings of human complex traits, with significant implications for genetic research.
• Projected sample sizes for identifying major heritability-explaining SNPs range from hundreds of thousands to millions, contingent on trait-specific effect-size distributions.
• The study provides insights into the potential and limitations of polygenic risk scores for disease prediction across different trait types.